{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "portfolio-introduction",
   "metadata": {},
   "source": "# Movie reviews to decision support\n## An explainable sentiment analysis portfolio\n\n**Gareth Grant · Marketing, analytics and machine learning**\n\nThis project asks whether movie-review text can distinguish positive from negative reviews,\nwhich method earns its complexity, and how the workflow could support a real organisation.\nIt rebuilds three supplied PDAN 8412 notebooks into one teaching notebook.\n\n**Reading pattern:** each numbered step has one purpose, one small code cell, and a guide to\nthe result. Add your own comment below a code cell or create a Markdown cell beside it.\nStable cell IDs also make individual steps easy to link from the HTML edition.\n\n**Execution status:** the complete classical workflow has been executed on the supplied CSV.\nThe optional TensorFlow pooling, SimpleRNN and LSTM experiments are provided as code, but\nwere **not trained or runtime-tested in this edition** because TensorFlow was unavailable\nand its installation could not complete. No neural performance is claimed.\n\n**Skills demonstrated:** data quality, leakage prevention, NLP, model comparison, validation,\nuncertainty, error analysis, explainability, reproducible exports and decision framing.\n\n### Contents\n\n1. [Question and study design](#question-and-study-design)\n2. [Setup](#setup)\n3. [Data audit and cleaning](#data-audit-and-cleaning)\n4. [Split before learning](#split-before-learning)\n5. [Explore the training data](#explore-the-training-data)\n6. [Turn text into features](#turn-text-into-features)\n7. [Classical models](#classical-models)\n8. [Optional neural experiments](#optional-neural-experiments)\n9. [Lock the model and operating rule](#lock-the-model-and-operating-rule)\n10. [Independent final evaluation](#independent-final-evaluation)\n11. [Explain and inspect errors](#explain-and-inspect-errors)\n12. [Predict new reviews](#predict-new-reviews)\n13. [Business, astronomy and government](#business-astronomy-and-government)\n14. [Reproducibility and publishing](#reproducibility-and-publishing)\n15. [Practice and references](#practice-and-references)\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-001",
   "metadata": {},
   "source": "## Question and study design\n\n**Question:** how reliably can a model classify strongly polarised English movie reviews,\nand what does it miss when used to flag negative feedback?\n\n| Item | Choice | Reason |\n|---|---|---|\n| Input | Review text only | Sentiment must not be supplied as a feature. |\n| Target | Negative = 0; positive = 1 | A binary task with an explicit label definition. |\n| Split | 60% train, 20% validation, 20% test | Learn, choose, then assess on separate rows. |\n| Model selection | Validation macro F1; log loss breaks ties | Treat both classes equally and consider probability quality. |\n| Default classification | Positive if score is at least 0.5 | A common comparison rule, locked before testing. |\n| Example triage rule | Tune a negative-score threshold on validation only | Show the trade-off between missed negatives and false alerts. |\n| Final metrics | Accuracy, macro F1, negative precision/recall, ROC AUC, log loss, Brier score | Different errors and probabilities need different measures. |\n\nThe source benchmark is the [Large Movie Review Dataset](https://ai.stanford.edu/~amaas/data/sentiment/)\nintroduced by [Maas et al. (2011)](https://aclanthology.org/P11-1015/). It contains\nstrong positive/negative reviews rather than a neutral class. The supplied combined CSV\nhas no original split, movie ID, author ID or date. Our new random split is therefore a\n**local evaluation, not the official Stanford benchmark split**. It cannot rule out shared\nfilms, authors or near-duplicate wording across partitions.\n\n### What changed from the original notebooks\n\n| Original issue | Portfolio improvement |\n|---|---|\n| The test set was repeatedly used for early stopping, activation comparison and tuning. | An independent validation set is used for those decisions. |\n| An embedding + average-pooling + Dense(ReLU) model was labelled an RNN. | It is correctly named a pooling network; SimpleRNN and LSTM are separate architectures. |\n| Duplicates were not excluded before splitting. | Audit raw and normalised duplicates, quarantine label conflicts, then deduplicate. |\n| Task 1 learned a tokenizer before splitting and kept text target labels. | Fit vocabulary on training only; encode targets explicitly. |\n| Task 1 referenced missing imports/callbacks and an absent GloVe file. | Self-contained core; optional neural section has explicit dependencies and needs no GloVe download. |\n| Tuner output reloaded old trials, while code requested one trial. | Use an explicit small validation search with recorded candidates; no hidden tuner cache. |\n| Early stopping did not restore the best weights. | Optional neural callbacks restore the best validation-loss weights. |\n| Already balanced data was sampled down and comments disagreed with sample size. | Use every valid unique review; report actual counts. |\n| Long paragraphs repeated metrics and mixed progress-bar estimates with full-set scores. | Calculate interpretations from the same stored predictions used for the metrics. |\n| No majority or strong classical baseline. | Include a majority baseline, Naive Bayes and tuned TF-IDF logistic regression. |\n\nSpark is useful when scale or a distributed platform justifies it. This 66 MB CSV can be\nhandled directly in pandas, which keeps the teaching workflow simpler and avoids the\noriginal CSV-parsing/Java dependencies.\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-002",
   "metadata": {},
   "source": "## Setup\n\nUse **Python 3.12** and install `requirements.txt` from the package. Launch JupyterLab,\nplace `IMDB Dataset.csv` beside this notebook, and choose **Restart Kernel and Run All**.\nChange only `DATA_PATH` if your data lives elsewhere. The package does not redistribute\nthe review dataset; keep your supplied CSV or obtain it from the linked dataset source.\n\nThe default run is complete without TensorFlow. For the optional experiments, install\n`requirements-neural.txt`, set `RUN_NEURAL = True`, and run all cells again. Neural scores\nwill then enter the same validation comparison before the final test section. Avoid\nchoosing a new model after looking at test results; further experiments need a new holdout.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-001-import-file-and-text-utilities-guide",
   "metadata": {},
   "source": "### 01. Import file and text utilities\n\nThese standard-library modules handle paths, normalisation and a reproducibility fingerprint.\n"
  },
  {
   "cell_type": "code",
   "id": "step-001-import-file-and-text-utilities",
   "metadata": {
    "portfolio_step": 1,
    "purpose": "Import file and text utilities",
    "execution_scope": "core",
    "execution_seconds": 0.0678
   },
   "execution_count": 1,
   "outputs": [],
   "source": "# Import file and text utilities\nfrom pathlib import Path\nimport hashlib\nimport html\nimport json\nimport re\nimport unicodedata\nimport platform\nimport importlib.metadata\nfrom time import perf_counter\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-002-import-data-and-plotting-libraries-guide",
   "metadata": {},
   "source": "### 02. Import data and plotting libraries\n\nKeep imports grouped by their job. `display` is provided by Jupyter.\n"
  },
  {
   "cell_type": "code",
   "id": "step-002-import-data-and-plotting-libraries",
   "metadata": {
    "portfolio_step": 2,
    "purpose": "Import data and plotting libraries",
    "execution_scope": "core",
    "execution_seconds": 1.9312
   },
   "execution_count": 2,
   "outputs": [],
   "source": "# Import data and plotting libraries\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport joblib\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-003-import-feature-and-model-tools-guide",
   "metadata": {},
   "source": "### 03. Import feature and model tools\n\nTF-IDF measures term importance. The two classifiers provide inexpensive, reproducible comparisons.\n"
  },
  {
   "cell_type": "code",
   "id": "step-003-import-feature-and-model-tools",
   "metadata": {
    "portfolio_step": 3,
    "purpose": "Import feature and model tools",
    "execution_scope": "core",
    "execution_seconds": 1.4787
   },
   "execution_count": 3,
   "outputs": [],
   "source": "# Import feature and model tools\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.feature_extraction.text import TfidfVectorizer\nfrom sklearn.dummy import DummyClassifier\nfrom sklearn.naive_bayes import MultinomialNB\nfrom sklearn.linear_model import LogisticRegression\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-004-import-evaluation-tools-guide",
   "metadata": {},
   "source": "### 04. Import evaluation tools\n\nEach metric answers a different question; accuracy alone can conceal costly errors.\n"
  },
  {
   "cell_type": "code",
   "id": "step-004-import-evaluation-tools",
   "metadata": {
    "portfolio_step": 4,
    "purpose": "Import evaluation tools",
    "execution_scope": "core",
    "execution_seconds": 0.0033
   },
   "execution_count": 4,
   "outputs": [],
   "source": "# Import evaluation tools\nfrom sklearn.metrics import (\n    accuracy_score, balanced_accuracy_score, f1_score,\n    precision_score, recall_score, roc_auc_score,\n    log_loss, brier_score_loss, confusion_matrix, classification_report,\n    RocCurveDisplay, PrecisionRecallDisplay,\n)\nfrom sklearn.calibration import calibration_curve\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-005-set-the-data-and-output-paths-guide",
   "metadata": {},
   "source": "### 05. Set the data and output paths\n\nAll output paths are relative to this project. Source data is never overwritten.\n"
  },
  {
   "cell_type": "code",
   "id": "step-005-set-the-data-and-output-paths",
   "metadata": {
    "portfolio_step": 5,
    "purpose": "Set the data and output paths",
    "execution_scope": "core",
    "execution_seconds": 0.0003
   },
   "execution_count": 5,
   "outputs": [],
   "source": "# Set the data and output paths\nDATA_PATH = Path('IMDB Dataset.csv')\nOUTPUT_DIR = Path('artifacts')\nOUTPUT_DIR.mkdir(exist_ok=True)\nif not DATA_PATH.is_file():\n    raise FileNotFoundError('Place IMDB Dataset.csv beside the notebook or edit DATA_PATH.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-006-set-reproducible-experiment-choices-guide",
   "metadata": {},
   "source": "### 06. Set reproducible experiment choices\n\nThese choices are fixed before examining the test set. Keep `RUN_NEURAL = False` for the tested core.\n"
  },
  {
   "cell_type": "code",
   "id": "step-006-set-reproducible-experiment-choices",
   "metadata": {
    "portfolio_step": 6,
    "purpose": "Set reproducible experiment choices",
    "execution_scope": "core",
    "execution_seconds": 0.0008
   },
   "execution_count": 6,
   "outputs": [],
   "source": "# Set reproducible experiment choices\nRANDOM_STATE = 42\nTFIDF_MAX_FEATURES = 50_000\nLOGISTIC_C_VALUES = [0.25, 1.0, 4.0]\nBOOTSTRAP_REPEATS = 1_000\nRUN_NEURAL = False\nMAX_LENGTH = 256\nNEURAL_VOCAB_SIZE = 15_000\nNEURAL_EPOCHS = 10\nNEURAL_BATCH_SIZE = 128\nnp.random.seed(RANDOM_STATE)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-007-set-a-readable-visual-style-guide",
   "metadata": {},
   "source": "### 07. Set a readable visual style\n\nCharts use Gareth’s navy and scarlet palette with labels that explain the measure.\n"
  },
  {
   "cell_type": "code",
   "id": "step-007-set-a-readable-visual-style",
   "metadata": {
    "portfolio_step": 7,
    "purpose": "Set a readable visual style",
    "execution_scope": "core",
    "execution_seconds": 0.001
   },
   "execution_count": 7,
   "outputs": [],
   "source": "# Set a readable visual style\nNAVY, SCARLET, SLATE = '#16324F', '#FF2400', '#667582'\nplt.rcParams.update({\n    'figure.figsize': (8, 4), 'figure.dpi': 120,\n    'axes.spines.top': False, 'axes.spines.right': False,\n    'axes.titleweight': 'bold', 'axes.labelcolor': NAVY,\n    'text.color': NAVY, 'font.size': 10,\n})\npd.set_option('display.max_colwidth', 180)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-008-record-the-source-fingerprint-guide",
   "metadata": {},
   "source": "### 08. Record the source fingerprint\n\nA SHA-256 hash identifies the exact file used; a matching filename alone is insufficient.\n"
  },
  {
   "cell_type": "code",
   "id": "step-008-record-the-source-fingerprint",
   "metadata": {
    "portfolio_step": 8,
    "purpose": "Record the source fingerprint",
    "execution_scope": "core",
    "execution_seconds": 0.2925
   },
   "execution_count": 8,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Dataset SHA-256: dfc447764f82be365fa9c2beef4e8df89d3919e3da95f5088004797d79695aa2\nDataset size: 66.21 MB\n"
    }
   ],
   "source": "# Record the source fingerprint\nsource_sha256 = hashlib.sha256(DATA_PATH.read_bytes()).hexdigest()\nprint('Dataset SHA-256:', source_sha256)\nprint(f'Dataset size: {DATA_PATH.stat().st_size / 1_000_000:.2f} MB')\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-019",
   "metadata": {},
   "source": "## Data audit and cleaning\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-009-read-the-csv-guide",
   "metadata": {},
   "source": "### 09. Read the CSV\n\nThe CSV parser respects quoted text containing commas. Load once and preserve a source-row index.\n"
  },
  {
   "cell_type": "code",
   "id": "step-009-read-the-csv",
   "metadata": {
    "portfolio_step": 9,
    "purpose": "Read the CSV",
    "execution_scope": "core",
    "execution_seconds": 0.8028
   },
   "execution_count": 9,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Loaded 50,000 rows and 2 columns.\n"
    }
   ],
   "source": "# Read the CSV\nraw = pd.read_csv(DATA_PATH, dtype={'review': 'string', 'sentiment': 'string'})\nraw.index.name = 'source_row'\nprint(f'Loaded {len(raw):,} rows and {raw.shape[1]} columns.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-010-check-the-required-schema-guide",
   "metadata": {},
   "source": "### 10. Check the required schema\n\nStop on missing fields or missing values instead of silently treating invalid labels as negative.\n"
  },
  {
   "cell_type": "code",
   "id": "step-010-check-the-required-schema",
   "metadata": {
    "portfolio_step": 10,
    "purpose": "Check the required schema",
    "execution_scope": "core",
    "execution_seconds": 0.0164
   },
   "execution_count": 10,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>dtype</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>review</th>\n      <td>string[python]</td>\n    </tr>\n    <tr>\n      <th>sentiment</th>\n      <td>string[python]</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                    dtype\nreview     string[python]\nsentiment  string[python]"
     },
     "metadata": {}
    }
   ],
   "source": "# Check the required schema\nrequired_columns = {'review', 'sentiment'}\nif not required_columns.issubset(raw.columns):\n    raise ValueError(f'Required columns: {sorted(required_columns)}')\nif raw[list(required_columns)].isna().any().any():\n    raise ValueError('Review or sentiment is missing; investigate before modelling.')\ndisplay(raw.dtypes.rename('dtype').to_frame())\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-011-validate-the-target-labels-guide",
   "metadata": {},
   "source": "### 11. Validate the target labels\n\nAn unknown label is a data-quality issue. It must not automatically become class zero.\n"
  },
  {
   "cell_type": "code",
   "id": "step-011-validate-the-target-labels",
   "metadata": {
    "portfolio_step": 11,
    "purpose": "Validate the target labels",
    "execution_scope": "core",
    "execution_seconds": 0.0255
   },
   "execution_count": 11,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>reviews</th>\n    </tr>\n    <tr>\n      <th>sentiment</th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>positive</th>\n      <td>25000</td>\n    </tr>\n    <tr>\n      <th>negative</th>\n      <td>25000</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "           reviews\nsentiment         \npositive     25000\nnegative     25000"
     },
     "metadata": {}
    }
   ],
   "source": "# Validate the target labels\nraw['sentiment'] = raw['sentiment'].str.strip().str.lower()\nunexpected_labels = set(raw['sentiment'].unique()) - {'negative', 'positive'}\nif unexpected_labels:\n    raise ValueError(f'Unexpected sentiment labels: {unexpected_labels}')\ndisplay(raw['sentiment'].value_counts().rename('reviews').to_frame())\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-012-audit-raw-duplicates-guide",
   "metadata": {},
   "source": "### 12. Audit raw duplicates\n\nRecord duplicates before changing the text, so the cleaning effect remains visible.\n"
  },
  {
   "cell_type": "code",
   "id": "step-012-audit-raw-duplicates",
   "metadata": {
    "portfolio_step": 12,
    "purpose": "Audit raw duplicates",
    "execution_scope": "core",
    "execution_seconds": 0.6398
   },
   "execution_count": 12,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Exact duplicate rows: 418\nRepeated raw reviews: 418; conflicting raw texts: 0\n"
    }
   ],
   "source": "# Audit raw duplicates\nraw_duplicate_rows = int(raw.duplicated(['review', 'sentiment']).sum())\nraw_duplicate_texts = int(raw.duplicated('review').sum())\nraw_conflicts = int(raw.groupby('review')['sentiment'].nunique().gt(1).sum())\nprint(f'Exact duplicate rows: {raw_duplicate_rows:,}')\nprint(f'Repeated raw reviews: {raw_duplicate_texts:,}; conflicting raw texts: {raw_conflicts:,}')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-013-define-minimal-text-normalisation-guide",
   "metadata": {},
   "source": "### 13. Define minimal text normalisation\n\nRemove HTML and formatting noise. Retain negation, apostrophes and numbers; do not delete stopwords such as “not”. This deterministic function does not learn from labels.\n"
  },
  {
   "cell_type": "code",
   "id": "step-013-define-minimal-text-normalisation",
   "metadata": {
    "portfolio_step": 13,
    "purpose": "Define minimal text normalisation",
    "execution_scope": "core",
    "execution_seconds": 0.0011
   },
   "execution_count": 13,
   "outputs": [],
   "source": "# Define minimal text normalisation\ndef normalise_text(text):\n    text = unicodedata.normalize('NFKC', html.unescape(str(text)))\n    text = text.replace('’', \"'\").lower()\n    text = re.sub(r'<[^>]*>', ' ', text)\n    text = re.sub(r\"[^a-z0-9'\\s]\", ' ', text)\n    return re.sub(r'\\s+', ' ', text).strip()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-014-build-the-normalised-text-column-guide",
   "metadata": {},
   "source": "### 14. Build the normalised text column\n\nKeep the original text available for audit; models use only the new `text` column.\n"
  },
  {
   "cell_type": "code",
   "id": "step-014-build-the-normalised-text-column",
   "metadata": {
    "portfolio_step": 14,
    "purpose": "Build the normalised text column",
    "execution_scope": "core",
    "execution_seconds": 6.027
   },
   "execution_count": 14,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Example: one of the other reviewers has mentioned that after watching just 1 oz episode you'll be hooked they are right as this is exactly what happened with me the first thing that struck …\n"
    }
   ],
   "source": "# Build the normalised text column\ndata = raw.reset_index().copy()\ndata['text'] = data['review'].map(normalise_text)\nprint('Example:', data['text'].iloc[0][:180] + '…')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-015-find-empty-and-conflicting-normalised-reviews-guide",
   "metadata": {},
   "source": "### 15. Find empty and conflicting normalised reviews\n\nA conflict means identical model input has different targets. Quarantine every row in that group rather than choosing a convenient label.\n"
  },
  {
   "cell_type": "code",
   "id": "step-015-find-empty-and-conflicting-normalised-reviews",
   "metadata": {
    "portfolio_step": 15,
    "purpose": "Find empty and conflicting normalised reviews",
    "execution_scope": "core",
    "execution_seconds": 0.4233
   },
   "execution_count": 15,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Empty texts: 0\nConflicting text groups: 0; affected rows: 0\nRepeated normalised texts: 423\n"
    }
   ],
   "source": "# Find empty and conflicting normalised reviews\nempty_mask = data['text'].eq('')\nconflict_texts = data.groupby('text')['sentiment'].nunique()\nconflict_texts = conflict_texts[conflict_texts > 1].index\nconflict_mask = data['text'].isin(conflict_texts)\nnormalised_duplicate_rows = int(data.duplicated('text').sum())\nprint(f'Empty texts: {empty_mask.sum():,}')\nprint(f'Conflicting text groups: {len(conflict_texts):,}; affected rows: {conflict_mask.sum():,}')\nprint(f'Repeated normalised texts: {normalised_duplicate_rows:,}')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-016-quarantine-invalid-text-rows-guide",
   "metadata": {},
   "source": "### 16. Quarantine invalid text rows\n\nCounts stay separate from the duplicate-removal count so every excluded row can be reconciled.\n"
  },
  {
   "cell_type": "code",
   "id": "step-016-quarantine-invalid-text-rows",
   "metadata": {
    "portfolio_step": 16,
    "purpose": "Quarantine invalid text rows",
    "execution_scope": "core",
    "execution_seconds": 0.0204
   },
   "execution_count": 16,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Quarantined rows: 0\n"
    }
   ],
   "source": "# Quarantine invalid text rows\nquarantine_mask = empty_mask | conflict_mask\nquarantined_rows = int(quarantine_mask.sum())\neligible = data.loc[~quarantine_mask].copy()\nprint(f'Quarantined rows: {quarantined_rows:,}')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-017-remove-repeated-model-inputs-guide",
   "metadata": {},
   "source": "### 17. Remove repeated model inputs\n\nKeep the first occurrence of each eligible normalised text before making any split.\n"
  },
  {
   "cell_type": "code",
   "id": "step-017-remove-repeated-model-inputs",
   "metadata": {
    "portfolio_step": 17,
    "purpose": "Remove repeated model inputs",
    "execution_scope": "core",
    "execution_seconds": 0.0279
   },
   "execution_count": 17,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Removed 423 repeated reviews; retained 49,577 unique reviews.\n"
    }
   ],
   "source": "# Remove repeated model inputs\nduplicates_removed = int(eligible.duplicated('text').sum())\nclean = eligible.drop_duplicates('text', keep='first').copy()\nclean = clean.reset_index(drop=True)\nassert len(raw) == quarantined_rows + duplicates_removed + len(clean)\nprint(f'Removed {duplicates_removed:,} repeated reviews; retained {len(clean):,} unique reviews.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-018-encode-the-target-guide",
   "metadata": {},
   "source": "### 18. Encode the target\n\nUse explicit integer targets: negative is zero; positive is one.\n"
  },
  {
   "cell_type": "code",
   "id": "step-018-encode-the-target",
   "metadata": {
    "portfolio_step": 18,
    "purpose": "Encode the target",
    "execution_scope": "core",
    "execution_seconds": 0.0144
   },
   "execution_count": 18,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>unique_reviews</th>\n    </tr>\n    <tr>\n      <th>sentiment</th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>positive</th>\n      <td>24881</td>\n    </tr>\n    <tr>\n      <th>negative</th>\n      <td>24696</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "           unique_reviews\nsentiment                \npositive            24881\nnegative            24696"
     },
     "metadata": {}
    }
   ],
   "source": "# Encode the target\nlabel_map = {'negative': 0, 'positive': 1}\nclean['label'] = clean['sentiment'].map(label_map).astype('int8')\ndisplay(clean['sentiment'].value_counts().rename('unique_reviews').to_frame())\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-019-assign-a-stable-review-identifier-guide",
   "metadata": {},
   "source": "### 19. Assign a stable review identifier\n\nHash the normalised input to identify rows in exported splits without publishing their full text. These hashes are identifiers, not anonymisation for sensitive data.\n"
  },
  {
   "cell_type": "code",
   "id": "step-019-assign-a-stable-review-identifier",
   "metadata": {
    "portfolio_step": 19,
    "purpose": "Assign a stable review identifier",
    "execution_scope": "core",
    "execution_seconds": 0.2158
   },
   "execution_count": 19,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "All retained review IDs are unique.\n"
    }
   ],
   "source": "# Assign a stable review identifier\nclean['review_id'] = clean['text'].map(lambda text: hashlib.sha256(text.encode()).hexdigest())\nassert clean['review_id'].is_unique\nprint('All retained review IDs are unique.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-042",
   "metadata": {},
   "source": "**Interpretation:** the original CSV is already balanced, so artificial resampling is not\nneeded. Deduplication may produce a small imbalance, which stratification preserves.\nNormalisation also exposes reviews differing only in formatting. Near duplicates are not\ndetected by this exact-match audit and remain a limitation.\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-043",
   "metadata": {},
   "source": "## Split before learning\n\nThe validation set is used to choose settings and the review threshold. The test set stays\nsealed until both decisions are recorded. Deterministic cleaning and duplicate auditing\nprecede the split; vocabulary, IDF weights and model parameters do not.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-020-reserve-the-final-test-set-guide",
   "metadata": {},
   "source": "### 20. Reserve the final test set\n\nStratification preserves the label proportions. Reserve 20% of unique reviews.\n"
  },
  {
   "cell_type": "code",
   "id": "step-020-reserve-the-final-test-set",
   "metadata": {
    "portfolio_step": 20,
    "purpose": "Reserve the final test set",
    "execution_scope": "core",
    "execution_seconds": 0.0563
   },
   "execution_count": 20,
   "outputs": [],
   "source": "# Reserve the final test set\ndevelopment_df, test_df = train_test_split(\n    clean, test_size=0.20, random_state=RANDOM_STATE, stratify=clean['label'],\n)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-021-separate-training-and-validation-guide",
   "metadata": {},
   "source": "### 21. Separate training and validation\n\nOne quarter of the remaining 80% is 20% of the complete cleaned dataset.\n"
  },
  {
   "cell_type": "code",
   "id": "step-021-separate-training-and-validation",
   "metadata": {
    "portfolio_step": 21,
    "purpose": "Separate training and validation",
    "execution_scope": "core",
    "execution_seconds": 0.0614
   },
   "execution_count": 21,
   "outputs": [],
   "source": "# Separate training and validation\ntrain_df, val_df = train_test_split(\n    development_df, test_size=0.25, random_state=RANDOM_STATE,\n    stratify=development_df['label'],\n)\ntrain_df, val_df, test_df = [frame.reset_index(drop=True) for frame in (train_df, val_df, test_df)]\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-022-verify-partition-independence-guide",
   "metadata": {},
   "source": "### 22. Verify partition independence\n\nThese checks catch accidental row reuse and cross-partition copies of normalised reviews.\n"
  },
  {
   "cell_type": "code",
   "id": "step-022-verify-partition-independence",
   "metadata": {
    "portfolio_step": 22,
    "purpose": "Verify partition independence",
    "execution_scope": "core",
    "execution_seconds": 0.0247
   },
   "execution_count": 22,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "No normalised review overlaps train, validation or test.\n"
    }
   ],
   "source": "# Verify partition independence\ntrain_ids, val_ids, test_ids = [set(frame['review_id']) for frame in (train_df, val_df, test_df)]\nassert train_ids.isdisjoint(val_ids)\nassert train_ids.isdisjoint(test_ids)\nassert val_ids.isdisjoint(test_ids)\nassert len(train_df) + len(val_df) + len(test_df) == len(clean)\nprint('No normalised review overlaps train, validation or test.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-023-report-the-development-split-guide",
   "metadata": {},
   "source": "### 23. Report the development split\n\nOnly training and validation class distributions are inspected before the final evaluation.\n"
  },
  {
   "cell_type": "code",
   "id": "step-023-report-the-development-split",
   "metadata": {
    "portfolio_step": 23,
    "purpose": "Report the development split",
    "execution_scope": "core",
    "execution_seconds": 0.0119
   },
   "execution_count": 23,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>rows</th>\n      <th>negative</th>\n      <th>positive</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>training</th>\n      <td>29745</td>\n      <td>14817</td>\n      <td>14928</td>\n    </tr>\n    <tr>\n      <th>validation</th>\n      <td>9916</td>\n      <td>4940</td>\n      <td>4976</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "             rows  negative  positive\ntraining    29745     14817     14928\nvalidation   9916      4940      4976"
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Reserved test rows: 9,916; not used for feature fitting or selection.\n"
    }
   ],
   "source": "# Report the development split\ndevelopment_summary = pd.DataFrame({\n    'rows': [len(train_df), len(val_df)],\n    'negative': [train_df['label'].eq(0).sum(), val_df['label'].eq(0).sum()],\n    'positive': [train_df['label'].eq(1).sum(), val_df['label'].eq(1).sum()],\n}, index=['training', 'validation'])\ndisplay(development_summary)\nprint(f'Reserved test rows: {len(test_df):,}; not used for feature fitting or selection.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-024-create-the-target-arrays-guide",
   "metadata": {},
   "source": "### 24. Create the target arrays\n\nThe arrays use the same row order as their DataFrames.\n"
  },
  {
   "cell_type": "code",
   "id": "step-024-create-the-target-arrays",
   "metadata": {
    "portfolio_step": 24,
    "purpose": "Create the target arrays",
    "execution_scope": "core",
    "execution_seconds": 0.0005
   },
   "execution_count": 24,
   "outputs": [],
   "source": "# Create the target arrays\ny_train = train_df['label'].to_numpy()\ny_val = val_df['label'].to_numpy()\ny_test = test_df['label'].to_numpy()\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-054",
   "metadata": {},
   "source": "## Explore the training data\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-025-count-training-review-lengths-guide",
   "metadata": {},
   "source": "### 25. Count training review lengths\n\nLength uses whitespace tokens after cleaning. These are descriptive counts, not the TF-IDF vocabulary.\n"
  },
  {
   "cell_type": "code",
   "id": "step-025-count-training-review-lengths",
   "metadata": {
    "portfolio_step": 25,
    "purpose": "Count training review lengths",
    "execution_scope": "core",
    "execution_seconds": 1.3685
   },
   "execution_count": 25,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>words</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>count</th>\n      <td>29745.0</td>\n    </tr>\n    <tr>\n      <th>mean</th>\n      <td>231.2</td>\n    </tr>\n    <tr>\n      <th>std</th>\n      <td>170.8</td>\n    </tr>\n    <tr>\n      <th>min</th>\n      <td>9.0</td>\n    </tr>\n    <tr>\n      <th>50%</th>\n      <td>174.0</td>\n    </tr>\n    <tr>\n      <th>90%</th>\n      <td>448.6</td>\n    </tr>\n    <tr>\n      <th>95%</th>\n      <td>586.0</td>\n    </tr>\n    <tr>\n      <th>max</th>\n      <td>2473.0</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "         words\ncount  29745.0\nmean     231.2\nstd      170.8\nmin        9.0\n50%      174.0\n90%      448.6\n95%      586.0\nmax     2473.0"
     },
     "metadata": {}
    }
   ],
   "source": "# Count training review lengths\ntrain_word_count = train_df['text'].str.split().str.len()\ndisplay(train_word_count.describe(percentiles=[0.50, 0.90, 0.95]).round(1).to_frame('words'))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-026-plot-the-training-class-balance-guide",
   "metadata": {},
   "source": "### 26. Plot the training class balance\n\nCount bars show the actual data used to fit the models.\n"
  },
  {
   "cell_type": "code",
   "id": "step-026-plot-the-training-class-balance",
   "metadata": {
    "portfolio_step": 26,
    "purpose": "Plot the training class balance",
    "execution_scope": "core",
    "execution_seconds": 0.7914
   },
   "execution_count": 26,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Plot the training class balance\nfig, ax = plt.subplots(figsize=(6, 3.5))\ncounts = train_df['sentiment'].value_counts().reindex(['negative', 'positive'])\nax.bar(counts.index, counts.values, color=[SCARLET, NAVY])\nax.set(title='Training reviews by sentiment', ylabel='Number of unique reviews')\nax.bar_label(ax.containers[0], fmt='%d', padding=3)\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-027-plot-the-review-length-distribution-guide",
   "metadata": {},
   "source": "### 27. Plot the review length distribution\n\nLong documents matter because optional sequence models truncate them. Classical models use the entire normalised review.\n"
  },
  {
   "cell_type": "code",
   "id": "step-027-plot-the-review-length-distribution",
   "metadata": {
    "portfolio_step": 27,
    "purpose": "Plot the review length distribution",
    "execution_scope": "core",
    "execution_seconds": 0.3231
   },
   "execution_count": 27,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Plot the review length distribution\nfig, ax = plt.subplots()\nax.hist(train_word_count, bins=60, color=NAVY)\nax.axvline(MAX_LENGTH, color=SCARLET, label=f'Neural limit: {MAX_LENGTH} words')\nax.set(xlim=(0, 1_000), xlabel='Words after cleaning', ylabel='Training reviews',\n       title='Training review lengths (view limited to 1,000 words)')\nax.legend()\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-028-measure-the-planned-truncation-rate-guide",
   "metadata": {},
   "source": "### 28. Measure the planned truncation rate\n\nQuantify the compromise rather than assuming 100 or 256 words preserve the whole review.\n"
  },
  {
   "cell_type": "code",
   "id": "step-028-measure-the-planned-truncation-rate",
   "metadata": {
    "portfolio_step": 28,
    "purpose": "Measure the planned truncation rate",
    "execution_scope": "core",
    "execution_seconds": 0.001
   },
   "execution_count": 28,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "28.6% of training reviews exceed the optional 256-word limit.\nSequence models would discard later words; TF-IDF does not impose this length limit.\n"
    }
   ],
   "source": "# Measure the planned truncation rate\ntruncation_rate = float(train_word_count.gt(MAX_LENGTH).mean())\nprint(f'{truncation_rate:.1%} of training reviews exceed the optional {MAX_LENGTH}-word limit.')\nprint('Sequence models would discard later words; TF-IDF does not impose this length limit.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-063",
   "metadata": {},
   "source": "## Turn text into features\n\n**Term frequency** measures how strongly a term appears in a review. **Inverse document\nfrequency** reduces the influence of terms appearing in many training reviews. A bigram\nis a two-word phrase such as “not good”, so the feature representation can capture some\nlocal word order. It cannot understand the complete sentence or sarcasm.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-029-define-the-tf-idf-representation-guide",
   "metadata": {},
   "source": "### 29. Define the TF-IDF representation\n\nSet all preprocessing choices explicitly. Retaining common words helps preserve negation.\n"
  },
  {
   "cell_type": "code",
   "id": "step-029-define-the-tf-idf-representation",
   "metadata": {
    "portfolio_step": 29,
    "purpose": "Define the TF-IDF representation",
    "execution_scope": "core",
    "execution_seconds": 0.0002
   },
   "execution_count": 29,
   "outputs": [],
   "source": "# Define the TF-IDF representation\nvectorizer = TfidfVectorizer(\n    max_features=TFIDF_MAX_FEATURES, ngram_range=(1, 2), min_df=2,\n    sublinear_tf=True, lowercase=False, stop_words=None,\n    token_pattern=r\"(?u)\\b\\w[\\w']+\\b\", dtype=np.float32,\n)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-030-fit-features-on-training-reviews-only-guide",
   "metadata": {},
   "source": "### 30. Fit features on training reviews only\n\nOnly this cell learns the vocabulary and IDF weights. Fitting on validation or test text would leak information.\n"
  },
  {
   "cell_type": "code",
   "id": "step-030-fit-features-on-training-reviews-only",
   "metadata": {
    "portfolio_step": 30,
    "purpose": "Fit features on training reviews only",
    "execution_scope": "core",
    "execution_seconds": 21.8837
   },
   "execution_count": 30,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Training matrix: (29745, 50000); nonzero entries: 7,093,508\nFeature fitting took 21.88 seconds on this machine.\n"
    }
   ],
   "source": "# Fit features on training reviews only\nfeature_start = perf_counter()\nX_train = vectorizer.fit_transform(train_df['text'])\nfeature_fit_seconds = perf_counter() - feature_start\nprint(f'Training matrix: {X_train.shape}; nonzero entries: {X_train.nnz:,}')\nprint(f'Feature fitting took {feature_fit_seconds:.2f} seconds on this machine.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-031-transform-validation-reviews-guide",
   "metadata": {},
   "source": "### 31. Transform validation reviews\n\nUse the training vocabulary without calling `fit` again. The test matrix is created later.\n"
  },
  {
   "cell_type": "code",
   "id": "step-031-transform-validation-reviews",
   "metadata": {
    "portfolio_step": 31,
    "purpose": "Transform validation reviews",
    "execution_scope": "core",
    "execution_seconds": 4.2243
   },
   "execution_count": 31,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Validation matrix: (9916, 50000)\n"
    }
   ],
   "source": "# Transform validation reviews\nX_val = vectorizer.transform(val_df['text'])\nassert X_train.shape[1] == X_val.shape[1]\nprint('Validation matrix:', X_val.shape)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-032-inspect-a-few-vocabulary-features-guide",
   "metadata": {},
   "source": "### 32. Inspect a few vocabulary features\n\nThese are learned features, not rankings of causally important words.\n"
  },
  {
   "cell_type": "code",
   "id": "step-032-inspect-a-few-vocabulary-features",
   "metadata": {
    "portfolio_step": 32,
    "purpose": "Inspect a few vocabulary features",
    "execution_scope": "core",
    "execution_seconds": 0.0696
   },
   "execution_count": 32,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>example_feature</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>00</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>000</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>000 000</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>000 to</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>007</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>01</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>02</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>10</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>10 000</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>10 10</td>\n    </tr>\n    <tr>\n      <th>10</th>\n      <td>10 and</td>\n    </tr>\n    <tr>\n      <th>11</th>\n      <td>10 as</td>\n    </tr>\n    <tr>\n      <th>12</th>\n      <td>10 because</td>\n    </tr>\n    <tr>\n      <th>13</th>\n      <td>10 but</td>\n    </tr>\n    <tr>\n      <th>14</th>\n      <td>10 for</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "   example_feature\n0               00\n1              000\n2          000 000\n3           000 to\n4              007\n5               01\n6               02\n7               10\n8           10 000\n9            10 10\n10          10 and\n11           10 as\n12      10 because\n13          10 but\n14          10 for"
     },
     "metadata": {}
    }
   ],
   "source": "# Inspect a few vocabulary features\nfeature_names = vectorizer.get_feature_names_out()\ndisplay(pd.DataFrame({'example_feature': feature_names[:15]}))\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-072",
   "metadata": {},
   "source": "## Classical models\n\nThe majority baseline ignores the review. Naive Bayes is a fast probabilistic text\nclassifier with a simplifying feature-independence assumption. Logistic regression learns\nweighted evidence for the two classes and is often a useful interpretable text baseline.\nHere only its regularisation strength is tuned, over three declared candidates.\n\n**Metric guide:** accuracy is the fraction correct; macro F1 gives each class equal weight;\nnegative recall measures how many negative reviews are caught; negative precision measures\nhow many flagged reviews are actually negative. ROC AUC measures ranking, while log loss\nand Brier score assess probability quality. Lower loss/Brier is better.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-033-define-one-consistent-metric-summary-guide",
   "metadata": {},
   "source": "### 33. Define one consistent metric summary\n\nEvery comparison uses the same label convention and 0.5 decision rule.\n"
  },
  {
   "cell_type": "code",
   "id": "step-033-define-one-consistent-metric-summary",
   "metadata": {
    "portfolio_step": 33,
    "purpose": "Define one consistent metric summary",
    "execution_scope": "core",
    "execution_seconds": 0.0004
   },
   "execution_count": 33,
   "outputs": [],
   "source": "# Define one consistent metric summary\ndef metric_summary(name, truth, positive_score):\n    prediction = (np.asarray(positive_score) >= 0.5).astype(int)\n    return {\n        'model': name,\n        'accuracy': accuracy_score(truth, prediction),\n        'balanced_accuracy': balanced_accuracy_score(truth, prediction),\n        'macro_f1': f1_score(truth, prediction, average='macro', zero_division=0),\n        'negative_precision': precision_score(truth, prediction, pos_label=0, zero_division=0),\n        'negative_recall': recall_score(truth, prediction, pos_label=0, zero_division=0),\n        'roc_auc': roc_auc_score(truth, positive_score),\n        'log_loss': log_loss(truth, positive_score, labels=[0, 1]),\n        'brier_score': brier_score_loss(truth, positive_score),\n    }\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-034-fit-the-majority-baseline-guide",
   "metadata": {},
   "source": "### 34. Fit the majority baseline\n\nA complex model must earn an improvement over predicting the training majority class.\n"
  },
  {
   "cell_type": "code",
   "id": "step-034-fit-the-majority-baseline",
   "metadata": {
    "portfolio_step": 34,
    "purpose": "Fit the majority baseline",
    "execution_scope": "core",
    "execution_seconds": 0.0407
   },
   "execution_count": 34,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>model</th>\n      <th>accuracy</th>\n      <th>balanced_accuracy</th>\n      <th>macro_f1</th>\n      <th>negative_precision</th>\n      <th>negative_recall</th>\n      <th>roc_auc</th>\n      <th>log_loss</th>\n      <th>brier_score</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>Majority baseline</td>\n      <td>0.5018</td>\n      <td>0.5</td>\n      <td>0.3341</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.6931</td>\n      <td>0.25</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "               model  accuracy  ...  log_loss  brier_score\n0  Majority baseline    0.5018  ...    0.6931         0.25\n\n[1 rows x 9 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Fit the majority baseline\nmajority_model = DummyClassifier(strategy='prior', random_state=RANDOM_STATE)\nmajority_model.fit(X_train, y_train)\nmajority_val_score = majority_model.predict_proba(X_val)[:, 1]\ndisplay(pd.DataFrame([metric_summary('Majority baseline', y_val, majority_val_score)]).round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-035-fit-the-naive-bayes-baseline-guide",
   "metadata": {},
   "source": "### 35. Fit the Naive Bayes baseline\n\nThe classifier receives the same training TF-IDF features as logistic regression.\n"
  },
  {
   "cell_type": "code",
   "id": "step-035-fit-the-naive-bayes-baseline",
   "metadata": {
    "portfolio_step": 35,
    "purpose": "Fit the Naive Bayes baseline",
    "execution_scope": "core",
    "execution_seconds": 0.1057
   },
   "execution_count": 35,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>model</th>\n      <th>accuracy</th>\n      <th>balanced_accuracy</th>\n      <th>macro_f1</th>\n      <th>negative_precision</th>\n      <th>negative_recall</th>\n      <th>roc_auc</th>\n      <th>log_loss</th>\n      <th>brier_score</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>TF-IDF + Naive Bayes</td>\n      <td>0.8858</td>\n      <td>0.8858</td>\n      <td>0.8858</td>\n      <td>0.8861</td>\n      <td>0.8846</td>\n      <td>0.9541</td>\n      <td>0.3192</td>\n      <td>0.0938</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                  model  accuracy  ...  log_loss  brier_score\n0  TF-IDF + Naive Bayes    0.8858  ...    0.3192       0.0938\n\n[1 rows x 9 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Fit the Naive Bayes baseline\nstart = perf_counter()\nnb_model = MultinomialNB(alpha=1.0).fit(X_train, y_train)\nnb_fit_seconds = perf_counter() - start\nnb_val_score = nb_model.predict_proba(X_val)[:, 1]\ndisplay(pd.DataFrame([metric_summary('TF-IDF + Naive Bayes', y_val, nb_val_score)]).round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-036-fit-the-declared-logistic-candidates-guide",
   "metadata": {},
   "source": "### 36. Fit the declared logistic candidates\n\nSmaller C means stronger regularisation. Evaluate candidates on validation only.\n"
  },
  {
   "cell_type": "code",
   "id": "step-036-fit-the-declared-logistic-candidates",
   "metadata": {
    "portfolio_step": 36,
    "purpose": "Fit the declared logistic candidates",
    "execution_scope": "core",
    "execution_seconds": 6.2552
   },
   "execution_count": 36,
   "outputs": [],
   "source": "# Fit the declared logistic candidates\nlogistic_models, tuning_rows = {}, []\nfor c_value in LOGISTIC_C_VALUES:\n    candidate = LogisticRegression(C=c_value, solver='liblinear', max_iter=300,\n                                   random_state=RANDOM_STATE)\n    start = perf_counter()\n    candidate.fit(X_train, y_train)\n    score = candidate.predict_proba(X_val)[:, 1]\n    row = metric_summary(f'C={c_value}', y_val, score)\n    row.update(C=c_value, fit_seconds=perf_counter() - start)\n    logistic_models[c_value] = candidate\n    tuning_rows.append(row)\ntuning_results = pd.DataFrame(tuning_rows)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-037-choose-logistic-regularisation-on-validation-guide",
   "metadata": {},
   "source": "### 37. Choose logistic regularisation on validation\n\nSelect by macro F1, then lower log loss, then smaller C. No test score participates.\n"
  },
  {
   "cell_type": "code",
   "id": "step-037-choose-logistic-regularisation-on-validation",
   "metadata": {
    "portfolio_step": 37,
    "purpose": "Choose logistic regularisation on validation",
    "execution_scope": "core",
    "execution_seconds": 0.0282
   },
   "execution_count": 37,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>model</th>\n      <th>accuracy</th>\n      <th>balanced_accuracy</th>\n      <th>macro_f1</th>\n      <th>negative_precision</th>\n      <th>negative_recall</th>\n      <th>roc_auc</th>\n      <th>log_loss</th>\n      <th>brier_score</th>\n      <th>C</th>\n      <th>fit_seconds</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>2</th>\n      <td>C=4.0</td>\n      <td>0.9103</td>\n      <td>0.9103</td>\n      <td>0.9103</td>\n      <td>0.9160</td>\n      <td>0.9028</td>\n      <td>0.9713</td>\n      <td>0.2363</td>\n      <td>0.0686</td>\n      <td>4.00</td>\n      <td>2.7200</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>C=1.0</td>\n      <td>0.9049</td>\n      <td>0.9049</td>\n      <td>0.9049</td>\n      <td>0.9113</td>\n      <td>0.8964</td>\n      <td>0.9682</td>\n      <td>0.2898</td>\n      <td>0.0821</td>\n      <td>1.00</td>\n      <td>1.5248</td>\n    </tr>\n    <tr>\n      <th>0</th>\n      <td>C=0.25</td>\n      <td>0.8902</td>\n      <td>0.8901</td>\n      <td>0.8902</td>\n      <td>0.8992</td>\n      <td>0.8779</td>\n      <td>0.9594</td>\n      <td>0.3866</td>\n      <td>0.1131</td>\n      <td>0.25</td>\n      <td>2.0093</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "    model  accuracy  balanced_accuracy  ...  brier_score     C  fit_seconds\n2   C=4.0    0.9103             0.9103  ...       0.0686  4.00       2.7200\n1   C=1.0    0.9049             0.9049  ...       0.0821  1.00       1.5248\n0  C=0.25    0.8902             0.8901  ...       0.1131  0.25       2.0093\n\n[3 rows x 11 columns]"
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Locked logistic C: 4\n"
    }
   ],
   "source": "# Choose logistic regularisation on validation\ntuning_results = tuning_results.sort_values(['macro_f1', 'log_loss', 'C'],\n                                            ascending=[False, True, True])\nbest_c = float(tuning_results.iloc[0]['C'])\nlogistic_model = logistic_models[best_c]\nlogistic_val_score = logistic_model.predict_proba(X_val)[:, 1]\ndisplay(tuning_results.round(4))\nprint(f'Locked logistic C: {best_c:g}')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-038-register-the-classical-comparison-models-guide",
   "metadata": {},
   "source": "### 38. Register the classical comparison models\n\nKeep model objects and validation predictions separate from the held-out test results.\n"
  },
  {
   "cell_type": "code",
   "id": "step-038-register-the-classical-comparison-models",
   "metadata": {
    "portfolio_step": 38,
    "purpose": "Register the classical comparison models",
    "execution_scope": "core",
    "execution_seconds": 0.0287
   },
   "execution_count": 38,
   "outputs": [],
   "source": "# Register the classical comparison models\ncore_models = {\n    'Majority baseline': majority_model,\n    'TF-IDF + Naive Bayes': nb_model,\n    'TF-IDF + Logistic Regression': logistic_model,\n}\nvalidation_scores = {\n    name: model.predict_proba(X_val)[:, 1] for name, model in core_models.items()\n}\nneural_models, neural_histories = {}, {}\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-085",
   "metadata": {},
   "source": "## Optional neural experiments\n\nThis section preserves the deep-learning learning goals of the original work, with correct\narchitecture names and training-only vocabulary. It is **optional and unexecuted in the\nsupplied results**. The guards keep the tested core runnable without TensorFlow. Setting\n`RUN_NEURAL = True` requires the neural dependencies and a fresh complete run.\n\n| Architecture | What it learns | Important limitation |\n|---|---|---|\n| Embedding + average pooling | A learned average of token embeddings | Averaging discards sequence order. ReLU does not make it recurrent. |\n| Embedding + SimpleRNN | A hidden state updated in token order | Long dependencies can be difficult to learn. |\n| Embedding + LSTM | A recurrent state with gates controlling memory | More parameters and compute; superiority must be measured. |\n\nAll three use the same data, vocabulary, length cap, embedding size, epoch limit, batch size\nand early-stopping rule. Their parameter counts and capacity still differ. This is a small\nteaching comparison, not an exhaustive architecture search. See the official\n[TensorFlow RNN tutorial](https://www.tensorflow.org/text/tutorials/text_classification_rnn).\n\nAn embedding with `mask_zero=True` prevents padding from being treated as a real word.\nEach training call receives a fresh early-stopping callback with\n[`restore_best_weights=True`](https://keras.io/api/callbacks/early_stopping/).\nHyperparameter changes must use validation results; the test section remains downstream.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-039-report-whether-neural-training-is-enabled-guide",
   "metadata": {},
   "source": "### 39. Report whether neural training is enabled\n\nA disabled experiment has no score. It must never appear as a trained result.\n"
  },
  {
   "cell_type": "code",
   "id": "step-039-report-whether-neural-training-is-enabled",
   "metadata": {
    "portfolio_step": 39,
    "purpose": "Report whether neural training is enabled",
    "execution_scope": "core",
    "execution_seconds": 0.0003
   },
   "execution_count": 39,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Neural experiments: NOT RUN. Core comparisons below are fully executed.\n"
    }
   ],
   "source": "# Report whether neural training is enabled\ntf = None\nif RUN_NEURAL:\n    import tensorflow as tf\n    tf.keras.utils.set_random_seed(RANDOM_STATE)\n    print('TensorFlow version:', tf.__version__)\nelse:\n    print('Neural experiments: NOT RUN. Core comparisons below are fully executed.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-040-create-the-neural-text-vectorizer-guide",
   "metadata": {},
   "source": "### 40. Create the neural text vectorizer\n\nToken zero is padding and token one represents an unknown word. Normalisation was already applied, so no second standardisation is needed.\n"
  },
  {
   "cell_type": "code",
   "id": "step-040-create-the-neural-text-vectorizer",
   "metadata": {
    "portfolio_step": 40,
    "purpose": "Create the neural text vectorizer",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0001
   },
   "execution_count": 40,
   "outputs": [],
   "source": "# Create the neural text vectorizer\nif RUN_NEURAL:\n    neural_vectorizer = tf.keras.layers.TextVectorization(\n        max_tokens=NEURAL_VOCAB_SIZE, standardize=None,\n        output_mode='int', output_sequence_length=MAX_LENGTH,\n    )\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-041-learn-the-neural-vocabulary-from-training-only-guide",
   "metadata": {},
   "source": "### 41. Learn the neural vocabulary from training only\n\nAdapt on training reviews; never adapt on validation or test reviews.\n"
  },
  {
   "cell_type": "code",
   "id": "step-041-learn-the-neural-vocabulary-from-training-only",
   "metadata": {
    "portfolio_step": 41,
    "purpose": "Learn the neural vocabulary from training only",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0002
   },
   "execution_count": 41,
   "outputs": [],
   "source": "# Learn the neural vocabulary from training only\nif RUN_NEURAL:\n    vocabulary_data = tf.data.Dataset.from_tensor_slices(train_df['text'].to_numpy()).batch(256)\n    neural_vectorizer.adapt(vocabulary_data)\n    neural_vocab_size = len(neural_vectorizer.get_vocabulary())\n    print('Neural vocabulary size, including special tokens:', neural_vocab_size)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-042-encode-the-training-sequences-guide",
   "metadata": {},
   "source": "### 42. Encode the training sequences\n\nThe fixed length uses padding and truncation. All neural models receive the same integer sequences.\n"
  },
  {
   "cell_type": "code",
   "id": "step-042-encode-the-training-sequences",
   "metadata": {
    "portfolio_step": 42,
    "purpose": "Encode the training sequences",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0002
   },
   "execution_count": 42,
   "outputs": [],
   "source": "# Encode the training sequences\nif RUN_NEURAL:\n    train_tokens = neural_vectorizer(train_df['text'].to_numpy()).numpy()\n    print('Training token shape:', train_tokens.shape)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-043-encode-the-validation-sequences-guide",
   "metadata": {},
   "source": "### 43. Encode the validation sequences\n\nUse the existing vocabulary without updating it. Do not encode the test set yet.\n"
  },
  {
   "cell_type": "code",
   "id": "step-043-encode-the-validation-sequences",
   "metadata": {
    "portfolio_step": 43,
    "purpose": "Encode the validation sequences",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0001
   },
   "execution_count": 43,
   "outputs": [],
   "source": "# Encode the validation sequences\nif RUN_NEURAL:\n    val_tokens = neural_vectorizer(val_df['text'].to_numpy()).numpy()\n    print('Validation token shape:', val_tokens.shape)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-044-prepare-the-neural-training-batches-guide",
   "metadata": {},
   "source": "### 44. Prepare the neural training batches\n\nShuffle training rows with a fixed seed; validation rows keep their order.\n"
  },
  {
   "cell_type": "code",
   "id": "step-044-prepare-the-neural-training-batches",
   "metadata": {
    "portfolio_step": 44,
    "purpose": "Prepare the neural training batches",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0003
   },
   "execution_count": 44,
   "outputs": [],
   "source": "# Prepare the neural training batches\nif RUN_NEURAL:\n    neural_train_data = tf.data.Dataset.from_tensor_slices((train_tokens, y_train.astype('float32')))\n    neural_train_data = neural_train_data.shuffle(len(y_train), seed=RANDOM_STATE)\n    neural_train_data = neural_train_data.batch(NEURAL_BATCH_SIZE).prefetch(tf.data.AUTOTUNE)\n    neural_val_data = tf.data.Dataset.from_tensor_slices((val_tokens, y_val.astype('float32')))\n    neural_val_data = neural_val_data.batch(NEURAL_BATCH_SIZE).prefetch(tf.data.AUTOTUNE)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-045-define-a-fresh-neural-stopping-rule-guide",
   "metadata": {},
   "source": "### 45. Define a fresh neural stopping rule\n\nMonitor validation loss and restore the best epoch. Do not reuse callback state across models.\n"
  },
  {
   "cell_type": "code",
   "id": "step-045-define-a-fresh-neural-stopping-rule",
   "metadata": {
    "portfolio_step": 45,
    "purpose": "Define a fresh neural stopping rule",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0003
   },
   "execution_count": 45,
   "outputs": [],
   "source": "# Define a fresh neural stopping rule\nif RUN_NEURAL:\n    def neural_callbacks():\n        return [tf.keras.callbacks.EarlyStopping(\n            monitor='val_loss', patience=2, restore_best_weights=True,\n        )]\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-046-build-the-average-pooling-network-guide",
   "metadata": {},
   "source": "### 46. Build the average-pooling network\n\nThis is the correctly named version of the original “ReLU RNN”. Its embedding average has no recurrent state.\n"
  },
  {
   "cell_type": "code",
   "id": "step-046-build-the-average-pooling-network",
   "metadata": {
    "portfolio_step": 46,
    "purpose": "Build the average-pooling network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0006
   },
   "execution_count": 46,
   "outputs": [],
   "source": "# Build the average-pooling network\nif RUN_NEURAL:\n    tf.keras.utils.set_random_seed(RANDOM_STATE)\n    pool_model = tf.keras.Sequential([\n        tf.keras.Input(shape=(MAX_LENGTH,), dtype='int32'),\n        tf.keras.layers.Embedding(neural_vocab_size, 32, mask_zero=True),\n        tf.keras.layers.GlobalAveragePooling1D(),\n        tf.keras.layers.Dense(16, activation='relu'),\n        tf.keras.layers.Dropout(0.2),\n        tf.keras.layers.Dense(1, activation='sigmoid'),\n    ], name='embedding_average_pooling')\n    pool_model.summary()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-047-compile-the-pooling-network-guide",
   "metadata": {},
   "source": "### 47. Compile the pooling network\n\nBinary cross-entropy matches the zero/one target. Gradient clipping constrains unusually large updates.\n"
  },
  {
   "cell_type": "code",
   "id": "step-047-compile-the-pooling-network",
   "metadata": {
    "portfolio_step": 47,
    "purpose": "Compile the pooling network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0003
   },
   "execution_count": 47,
   "outputs": [],
   "source": "# Compile the pooling network\nif RUN_NEURAL:\n    pool_model.compile(optimizer=tf.keras.optimizers.Adam(0.001, clipnorm=1.0),\n                       loss='binary_crossentropy', metrics=['accuracy'])\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-048-train-the-pooling-network-guide",
   "metadata": {},
   "source": "### 48. Train the pooling network\n\nOnly training data updates weights; validation loss controls stopping.\n"
  },
  {
   "cell_type": "code",
   "id": "step-048-train-the-pooling-network",
   "metadata": {
    "portfolio_step": 48,
    "purpose": "Train the pooling network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0003
   },
   "execution_count": 48,
   "outputs": [],
   "source": "# Train the pooling network\nif RUN_NEURAL:\n    pool_history = pool_model.fit(neural_train_data, validation_data=neural_val_data,\n                                  epochs=NEURAL_EPOCHS, callbacks=neural_callbacks(), verbose=2)\n    neural_models['Embedding + average pooling'] = pool_model\n    neural_histories['Embedding + average pooling'] = pool_history.history\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-049-build-the-simplernn-network-guide",
   "metadata": {},
   "source": "### 49. Build the SimpleRNN network\n\nSimpleRNN is a genuine recurrent layer. Here tanh is the state activation, not a separate architecture name.\n"
  },
  {
   "cell_type": "code",
   "id": "step-049-build-the-simplernn-network",
   "metadata": {
    "portfolio_step": 49,
    "purpose": "Build the SimpleRNN network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0007
   },
   "execution_count": 49,
   "outputs": [],
   "source": "# Build the SimpleRNN network\nif RUN_NEURAL:\n    tf.keras.utils.set_random_seed(RANDOM_STATE)\n    rnn_model = tf.keras.Sequential([\n        tf.keras.Input(shape=(MAX_LENGTH,), dtype='int32'),\n        tf.keras.layers.Embedding(neural_vocab_size, 32, mask_zero=True),\n        tf.keras.layers.SimpleRNN(32, activation='tanh'),\n        tf.keras.layers.Dropout(0.2),\n        tf.keras.layers.Dense(1, activation='sigmoid'),\n    ], name='simple_rnn')\n    rnn_model.summary()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-050-compile-the-simplernn-network-guide",
   "metadata": {},
   "source": "### 50. Compile the SimpleRNN network\n\nUse the same loss and optimisation choices as the other neural models.\n"
  },
  {
   "cell_type": "code",
   "id": "step-050-compile-the-simplernn-network",
   "metadata": {
    "portfolio_step": 50,
    "purpose": "Compile the SimpleRNN network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0007
   },
   "execution_count": 50,
   "outputs": [],
   "source": "# Compile the SimpleRNN network\nif RUN_NEURAL:\n    rnn_model.compile(optimizer=tf.keras.optimizers.Adam(0.001, clipnorm=1.0),\n                      loss='binary_crossentropy', metrics=['accuracy'])\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-051-train-the-simplernn-network-guide",
   "metadata": {},
   "source": "### 51. Train the SimpleRNN network\n\nThe callback is created afresh for this run.\n"
  },
  {
   "cell_type": "code",
   "id": "step-051-train-the-simplernn-network",
   "metadata": {
    "portfolio_step": 51,
    "purpose": "Train the SimpleRNN network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0004
   },
   "execution_count": 51,
   "outputs": [],
   "source": "# Train the SimpleRNN network\nif RUN_NEURAL:\n    rnn_history = rnn_model.fit(neural_train_data, validation_data=neural_val_data,\n                                epochs=NEURAL_EPOCHS, callbacks=neural_callbacks(), verbose=2)\n    neural_models['Embedding + SimpleRNN'] = rnn_model\n    neural_histories['Embedding + SimpleRNN'] = rnn_history.history\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-052-build-the-lstm-network-guide",
   "metadata": {},
   "source": "### 52. Build the LSTM network\n\nLSTM adds memory gates. A more complex model must still outperform simpler baselines.\n"
  },
  {
   "cell_type": "code",
   "id": "step-052-build-the-lstm-network",
   "metadata": {
    "portfolio_step": 52,
    "purpose": "Build the LSTM network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0131
   },
   "execution_count": 52,
   "outputs": [],
   "source": "# Build the LSTM network\nif RUN_NEURAL:\n    tf.keras.utils.set_random_seed(RANDOM_STATE)\n    lstm_model = tf.keras.Sequential([\n        tf.keras.Input(shape=(MAX_LENGTH,), dtype='int32'),\n        tf.keras.layers.Embedding(neural_vocab_size, 32, mask_zero=True),\n        tf.keras.layers.LSTM(32),\n        tf.keras.layers.Dropout(0.2),\n        tf.keras.layers.Dense(1, activation='sigmoid'),\n    ], name='lstm')\n    lstm_model.summary()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-053-compile-the-lstm-network-guide",
   "metadata": {},
   "source": "### 53. Compile the LSTM network\n\nKeep the training objective consistent with the comparison models.\n"
  },
  {
   "cell_type": "code",
   "id": "step-053-compile-the-lstm-network",
   "metadata": {
    "portfolio_step": 53,
    "purpose": "Compile the LSTM network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0003
   },
   "execution_count": 53,
   "outputs": [],
   "source": "# Compile the LSTM network\nif RUN_NEURAL:\n    lstm_model.compile(optimizer=tf.keras.optimizers.Adam(0.001, clipnorm=1.0),\n                       loss='binary_crossentropy', metrics=['accuracy'])\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-054-train-the-lstm-network-guide",
   "metadata": {},
   "source": "### 54. Train the LSTM network\n\nRestore the best validation-loss weights rather than evaluating the final deteriorated epoch.\n"
  },
  {
   "cell_type": "code",
   "id": "step-054-train-the-lstm-network",
   "metadata": {
    "portfolio_step": 54,
    "purpose": "Train the LSTM network",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0003
   },
   "execution_count": 54,
   "outputs": [],
   "source": "# Train the LSTM network\nif RUN_NEURAL:\n    lstm_history = lstm_model.fit(neural_train_data, validation_data=neural_val_data,\n                                  epochs=NEURAL_EPOCHS, callbacks=neural_callbacks(), verbose=2)\n    neural_models['Embedding + LSTM'] = lstm_model\n    neural_histories['Embedding + LSTM'] = lstm_history.history\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-055-compare-neural-training-curves-guide",
   "metadata": {},
   "source": "### 55. Compare neural training curves\n\nFalling training loss alongside rising validation loss suggests overfitting. No curves are fabricated when this section is disabled.\n"
  },
  {
   "cell_type": "code",
   "id": "step-055-compare-neural-training-curves",
   "metadata": {
    "portfolio_step": 55,
    "purpose": "Compare neural training curves",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0004
   },
   "execution_count": 55,
   "outputs": [],
   "source": "# Compare neural training curves\nif RUN_NEURAL:\n    fig, ax = plt.subplots()\n    for name, history in neural_histories.items():\n        epochs = np.arange(1, len(history['loss']) + 1)\n        ax.plot(epochs, history['val_loss'], label=name)\n    ax.set(xlabel='Epoch', ylabel='Validation binary cross-entropy',\n           title='Neural validation learning curves')\n    ax.legend()\n    fig.tight_layout()\n    plt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-056-register-neural-validation-predictions-guide",
   "metadata": {},
   "source": "### 56. Register neural validation predictions\n\nNeural models enter selection only after actual training. Flatten output arrays before metric calculations.\n"
  },
  {
   "cell_type": "code",
   "id": "step-056-register-neural-validation-predictions",
   "metadata": {
    "portfolio_step": 56,
    "purpose": "Register neural validation predictions",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0002
   },
   "execution_count": 56,
   "outputs": [],
   "source": "# Register neural validation predictions\nif RUN_NEURAL:\n    for name, model in neural_models.items():\n        validation_scores[name] = model.predict(val_tokens, verbose=0).ravel()\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-122",
   "metadata": {},
   "source": "## Lock the model and operating rule\n\nThe comparison model is chosen on validation macro F1. Separately, an illustrative\nnegative-feedback queue uses a cost rule: **a missed negative costs five units; a false\nalert costs one unit**. These units are declared assumptions, not measured rand, customer\nlosses or benefits. A real organisation must estimate its own error costs and review capacity.\n\n`negative_score = 1 - positive_score`. Flag a review when its negative score reaches a\nchosen threshold. Lower thresholds usually catch more negatives and create more alerts.\nThe scores are model estimates; they are not established calibrated probabilities.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-057-compare-all-executed-models-on-validation-guide",
   "metadata": {},
   "source": "### 57. Compare all executed models on validation\n\nA model that was not trained is absent from this score table.\n"
  },
  {
   "cell_type": "code",
   "id": "step-057-compare-all-executed-models-on-validation",
   "metadata": {
    "portfolio_step": 57,
    "purpose": "Compare all executed models on validation",
    "execution_scope": "core",
    "execution_seconds": 0.076
   },
   "execution_count": 57,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>model</th>\n      <th>accuracy</th>\n      <th>balanced_accuracy</th>\n      <th>macro_f1</th>\n      <th>negative_precision</th>\n      <th>negative_recall</th>\n      <th>roc_auc</th>\n      <th>log_loss</th>\n      <th>brier_score</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>2</th>\n      <td>TF-IDF + Logistic Regression</td>\n      <td>0.9103</td>\n      <td>0.9103</td>\n      <td>0.9103</td>\n      <td>0.9160</td>\n      <td>0.9028</td>\n      <td>0.9713</td>\n      <td>0.2363</td>\n      <td>0.0686</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>TF-IDF + Naive Bayes</td>\n      <td>0.8858</td>\n      <td>0.8858</td>\n      <td>0.8858</td>\n      <td>0.8861</td>\n      <td>0.8846</td>\n      <td>0.9541</td>\n      <td>0.3192</td>\n      <td>0.0938</td>\n    </tr>\n    <tr>\n      <th>0</th>\n      <td>Majority baseline</td>\n      <td>0.5018</td>\n      <td>0.5000</td>\n      <td>0.3341</td>\n      <td>0.0000</td>\n      <td>0.0000</td>\n      <td>0.5000</td>\n      <td>0.6931</td>\n      <td>0.2500</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                          model  accuracy  ...  log_loss  brier_score\n2  TF-IDF + Logistic Regression    0.9103  ...    0.2363       0.0686\n1          TF-IDF + Naive Bayes    0.8858  ...    0.3192       0.0938\n0             Majority baseline    0.5018  ...    0.6931       0.2500\n\n[3 rows x 9 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Compare all executed models on validation\nvalidation_results = pd.DataFrame([\n    metric_summary(name, y_val, score) for name, score in validation_scores.items()\n]).sort_values(['macro_f1', 'log_loss', 'model'], ascending=[False, True, True])\ndisplay(validation_results.round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-058-lock-the-selected-model-guide",
   "metadata": {},
   "source": "### 58. Lock the selected model\n\nDo not change this choice after seeing the test results. Keep the strongest core model available for explanations and export.\n"
  },
  {
   "cell_type": "code",
   "id": "step-058-lock-the-selected-model",
   "metadata": {
    "portfolio_step": 58,
    "purpose": "Lock the selected model",
    "execution_scope": "core",
    "execution_seconds": 0.0014
   },
   "execution_count": 58,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Model selected on validation: TF-IDF + Logistic Regression\nBest classical model: TF-IDF + Logistic Regression\n"
    }
   ],
   "source": "# Lock the selected model\nselected_name = str(validation_results.iloc[0]['model'])\ncore_ranking = validation_results[validation_results['model'].isin(core_models)]\nbest_core_name = str(core_ranking.iloc[0]['model'])\nbest_core_model = core_models[best_core_name]\nprint('Model selected on validation:', selected_name)\nprint('Best classical model:', best_core_name)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-059-define-the-negative-feedback-cost-calculation-guide",
   "metadata": {},
   "source": "### 59. Define the negative-feedback cost calculation\n\nThe names refer to a negative-review queue: a false alert is a positive review sent for negative-feedback review.\n"
  },
  {
   "cell_type": "code",
   "id": "step-059-define-the-negative-feedback-cost-calculation",
   "metadata": {
    "portfolio_step": 59,
    "purpose": "Define the negative-feedback cost calculation",
    "execution_scope": "core",
    "execution_seconds": 0.0008
   },
   "execution_count": 59,
   "outputs": [],
   "source": "# Define the negative-feedback cost calculation\ndef triage_summary(truth, positive_score, threshold):\n    is_negative = np.asarray(truth) == 0\n    flagged = (1 - np.asarray(positive_score)) >= threshold\n    caught = int((is_negative & flagged).sum())\n    missed = int((is_negative & ~flagged).sum())\n    false_alerts = int((~is_negative & flagged).sum())\n    return {\n        'threshold': float(threshold), 'flagged': int(flagged.sum()),\n        'negative_recall': caught / max(int(is_negative.sum()), 1),\n        'negative_precision': caught / max(int(flagged.sum()), 1),\n        'missed_negatives': missed, 'false_alerts': false_alerts,\n        'relative_cost_units': 5 * missed + false_alerts,\n    }\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-060-evaluate-declared-thresholds-on-validation-guide",
   "metadata": {},
   "source": "### 60. Evaluate declared thresholds on validation\n\nThe threshold grid is specified before testing. No optimisation occurs on the test labels.\n"
  },
  {
   "cell_type": "code",
   "id": "step-060-evaluate-declared-thresholds-on-validation",
   "metadata": {
    "portfolio_step": 60,
    "purpose": "Evaluate declared thresholds on validation",
    "execution_scope": "core",
    "execution_seconds": 0.0134
   },
   "execution_count": 60,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>threshold</th>\n      <th>flagged</th>\n      <th>negative_recall</th>\n      <th>negative_precision</th>\n      <th>missed_negatives</th>\n      <th>false_alerts</th>\n      <th>relative_cost_units</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0.10</td>\n      <td>7290</td>\n      <td>0.9947</td>\n      <td>0.6741</td>\n      <td>26</td>\n      <td>2376</td>\n      <td>2506</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0.15</td>\n      <td>6700</td>\n      <td>0.9899</td>\n      <td>0.7299</td>\n      <td>50</td>\n      <td>1810</td>\n      <td>2060</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0.20</td>\n      <td>6276</td>\n      <td>0.9822</td>\n      <td>0.7731</td>\n      <td>88</td>\n      <td>1424</td>\n      <td>1864</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>0.25</td>\n      <td>5934</td>\n      <td>0.9739</td>\n      <td>0.8108</td>\n      <td>129</td>\n      <td>1123</td>\n      <td>1768</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>0.30</td>\n      <td>5689</td>\n      <td>0.9634</td>\n      <td>0.8365</td>\n      <td>181</td>\n      <td>930</td>\n      <td>1835</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>0.35</td>\n      <td>5484</td>\n      <td>0.9524</td>\n      <td>0.8580</td>\n      <td>235</td>\n      <td>779</td>\n      <td>1954</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>0.40</td>\n      <td>5264</td>\n      <td>0.9370</td>\n      <td>0.8794</td>\n      <td>311</td>\n      <td>635</td>\n      <td>2190</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>0.45</td>\n      <td>5064</td>\n      <td>0.9211</td>\n      <td>0.8985</td>\n      <td>390</td>\n      <td>514</td>\n      <td>2464</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>0.50</td>\n      <td>4869</td>\n      <td>0.9028</td>\n      <td>0.9160</td>\n      <td>480</td>\n      <td>409</td>\n      <td>2809</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>0.55</td>\n      <td>4686</td>\n      <td>0.8832</td>\n      <td>0.9311</td>\n      <td>577</td>\n      <td>323</td>\n      <td>3208</td>\n    </tr>\n    <tr>\n      <th>10</th>\n      <td>0.60</td>\n      <td>4498</td>\n      <td>0.8583</td>\n      <td>0.9426</td>\n      <td>700</td>\n      <td>258</td>\n      <td>3758</td>\n    </tr>\n    <tr>\n      <th>11</th>\n      <td>0.65</td>\n      <td>4314</td>\n      <td>0.8328</td>\n      <td>0.9536</td>\n      <td>826</td>\n      <td>200</td>\n      <td>4330</td>\n    </tr>\n    <tr>\n      <th>12</th>\n      <td>0.70</td>\n      <td>4098</td>\n      <td>0.8002</td>\n      <td>0.9646</td>\n      <td>987</td>\n      <td>145</td>\n      <td>5080</td>\n    </tr>\n    <tr>\n      <th>13</th>\n      <td>0.75</td>\n      <td>3871</td>\n      <td>0.7609</td>\n      <td>0.9711</td>\n      <td>1181</td>\n      <td>112</td>\n      <td>6017</td>\n    </tr>\n    <tr>\n      <th>14</th>\n      <td>0.80</td>\n      <td>3592</td>\n      <td>0.7109</td>\n      <td>0.9777</td>\n      <td>1428</td>\n      <td>80</td>\n      <td>7220</td>\n    </tr>\n    <tr>\n      <th>15</th>\n      <td>0.85</td>\n      <td>3216</td>\n      <td>0.6415</td>\n      <td>0.9854</td>\n      <td>1771</td>\n      <td>47</td>\n      <td>8902</td>\n    </tr>\n    <tr>\n      <th>16</th>\n      <td>0.90</td>\n      <td>2728</td>\n      <td>0.5484</td>\n      <td>0.9930</td>\n      <td>2231</td>\n      <td>19</td>\n      <td>11174</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "    threshold  flagged  ...  false_alerts  relative_cost_units\n0        0.10     7290  ...          2376                 2506\n1        0.15     6700  ...          1810                 2060\n2        0.20     6276  ...          1424                 1864\n3        0.25     5934  ...          1123                 1768\n4        0.30     5689  ...           930                 1835\n5        0.35     5484  ...           779                 1954\n6        0.40     5264  ...           635                 2190\n7        0.45     5064  ...           514                 2464\n8        0.50     4869  ...           409                 2809\n9        0.55     4686  ...           323                 3208\n10       0.60     4498  ...           258                 3758\n11       0.65     4314  ...           200                 4330\n12       0.70     4098  ...           145                 5080\n13       0.75     3871  ...           112                 6017\n14       0.80     3592  ...            80                 7220\n15       0.85     3216  ...            47                 8902\n16       0.90     2728  ...            19                11174\n\n[17 rows x 7 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Evaluate declared thresholds on validation\nthreshold_grid = np.round(np.arange(0.10, 0.91, 0.05), 2)\nthreshold_results = pd.DataFrame([\n    triage_summary(y_val, validation_scores[selected_name], threshold)\n    for threshold in threshold_grid\n])\ndisplay(threshold_results.round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-061-lock-the-illustrative-triage-threshold-guide",
   "metadata": {},
   "source": "### 61. Lock the illustrative triage threshold\n\nIf costs tie, prefer the value nearest 0.5. This policy is independent of the default 0.5 classification comparison.\n"
  },
  {
   "cell_type": "code",
   "id": "step-061-lock-the-illustrative-triage-threshold",
   "metadata": {
    "portfolio_step": 61,
    "purpose": "Lock the illustrative triage threshold",
    "execution_scope": "core",
    "execution_seconds": 0.0021
   },
   "execution_count": 61,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "{\n  \"selected_model\": \"TF-IDF + Logistic Regression\",\n  \"logistic_C\": 4.0,\n  \"classification_threshold_positive\": 0.5,\n  \"triage_threshold_negative\": 0.25,\n  \"missed_negative_cost_units\": 5,\n  \"false_alert_cost_units\": 1\n}\n"
    }
   ],
   "source": "# Lock the illustrative triage threshold\nthreshold_results['distance_from_half'] = (threshold_results['threshold'] - 0.5).abs()\nthreshold_choice = threshold_results.sort_values(\n    ['relative_cost_units', 'distance_from_half', 'threshold']\n).iloc[0]\ntriage_threshold = float(threshold_choice['threshold'])\nlocked_plan = {'selected_model': selected_name, 'logistic_C': best_c,\n               'classification_threshold_positive': 0.5,\n               'triage_threshold_negative': triage_threshold,\n               'missed_negative_cost_units': 5, 'false_alert_cost_units': 1}\nprint(json.dumps(locked_plan, indent=2))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-062-plot-the-validation-threshold-trade-off-guide",
   "metadata": {},
   "source": "### 62. Plot the validation threshold trade-off\n\nThe chart makes review workload and missed-negative cost visible. It is not a return-on-investment result.\n"
  },
  {
   "cell_type": "code",
   "id": "step-062-plot-the-validation-threshold-trade-off",
   "metadata": {
    "portfolio_step": 62,
    "purpose": "Plot the validation threshold trade-off",
    "execution_scope": "core",
    "execution_seconds": 0.3002
   },
   "execution_count": 62,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Plot the validation threshold trade-off\nfig, ax = plt.subplots()\nax.plot(threshold_results['threshold'], threshold_results['relative_cost_units'], color=NAVY)\nax.axvline(triage_threshold, color=SCARLET, label=f'Locked threshold: {triage_threshold:.2f}')\nax.set(xlabel='Threshold for negative-review flag', ylabel='Relative error-cost units',\n       title='Illustrative triage cost on validation only')\nax.legend()\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-135",
   "metadata": {},
   "source": "## Independent final evaluation\n\nAll decisions above are now fixed. The following test results describe this dataset and\nthis trained run. They do not estimate performance on product reviews, South African\nlanguages, citizen complaints or scientific papers. Validation scores are optimistic after\nselection; the held-out test is the final assessment of the locked choices.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-063-transform-the-sealed-test-reviews-guide",
   "metadata": {},
   "source": "### 63. Transform the sealed test reviews\n\nUse the already-fitted training vocabulary. There is no fit, adapt or tuning call here.\n"
  },
  {
   "cell_type": "code",
   "id": "step-063-transform-the-sealed-test-reviews",
   "metadata": {
    "portfolio_step": 63,
    "purpose": "Transform the sealed test reviews",
    "execution_scope": "core",
    "execution_seconds": 4.8029
   },
   "execution_count": 63,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Final test matrix: (9916, 50000)\n"
    }
   ],
   "source": "# Transform the sealed test reviews\nX_test = vectorizer.transform(test_df['text'])\nprint('Final test matrix:', X_test.shape)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-064-make-final-classical-predictions-guide",
   "metadata": {},
   "source": "### 64. Make final classical predictions\n\nStore each score vector once so tables, charts and interpretations use the same evidence.\n"
  },
  {
   "cell_type": "code",
   "id": "step-064-make-final-classical-predictions",
   "metadata": {
    "portfolio_step": 64,
    "purpose": "Make final classical predictions",
    "execution_scope": "core",
    "execution_seconds": 0.0335
   },
   "execution_count": 64,
   "outputs": [],
   "source": "# Make final classical predictions\ntest_scores, prediction_seconds = {}, {}\nfor name, model in core_models.items():\n    start = perf_counter()\n    test_scores[name] = model.predict_proba(X_test)[:, 1]\n    prediction_seconds[name] = perf_counter() - start\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-065-make-final-neural-predictions-when-enabled-guide",
   "metadata": {},
   "source": "### 65. Make final neural predictions when enabled\n\nThis runs only for actually trained neural models, using the frozen neural vocabulary.\n"
  },
  {
   "cell_type": "code",
   "id": "step-065-make-final-neural-predictions-when-enabled",
   "metadata": {
    "portfolio_step": 65,
    "purpose": "Make final neural predictions when enabled",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0005
   },
   "execution_count": 65,
   "outputs": [],
   "source": "# Make final neural predictions when enabled\nif RUN_NEURAL:\n    test_tokens = neural_vectorizer(test_df['text'].to_numpy()).numpy()\n    for name, model in neural_models.items():\n        start = perf_counter()\n        test_scores[name] = model.predict(test_tokens, verbose=0).ravel()\n        prediction_seconds[name] = perf_counter() - start\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-066-check-prediction-integrity-guide",
   "metadata": {},
   "source": "### 66. Check prediction integrity\n\nInvalid scores or wrong row counts should fail before a result can be published.\n"
  },
  {
   "cell_type": "code",
   "id": "step-066-check-prediction-integrity",
   "metadata": {
    "portfolio_step": 66,
    "purpose": "Check prediction integrity",
    "execution_scope": "core",
    "execution_seconds": 0.001
   },
   "execution_count": 66,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "All executed models returned one valid score per held-out review.\n"
    }
   ],
   "source": "# Check prediction integrity\nfor name, score in test_scores.items():\n    assert score.shape == y_test.shape, name\n    assert np.isfinite(score).all(), name\n    assert ((score >= 0) & (score <= 1)).all(), name\nprint('All executed models returned one valid score per held-out review.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-067-build-the-final-comparison-table-guide",
   "metadata": {},
   "source": "### 67. Build the final comparison table\n\nPrediction time excludes text cleaning and vectorisation; it is not a web-service latency measurement.\n"
  },
  {
   "cell_type": "code",
   "id": "step-067-build-the-final-comparison-table",
   "metadata": {
    "portfolio_step": 67,
    "purpose": "Build the final comparison table",
    "execution_scope": "core",
    "execution_seconds": 0.1952
   },
   "execution_count": 67,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>model</th>\n      <th>accuracy</th>\n      <th>balanced_accuracy</th>\n      <th>macro_f1</th>\n      <th>negative_precision</th>\n      <th>negative_recall</th>\n      <th>roc_auc</th>\n      <th>log_loss</th>\n      <th>brier_score</th>\n      <th>prediction_seconds_without_preprocessing</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>Majority baseline</td>\n      <td>0.5019</td>\n      <td>0.5000</td>\n      <td>0.3342</td>\n      <td>0.0000</td>\n      <td>0.0000</td>\n      <td>0.5000</td>\n      <td>0.6931</td>\n      <td>0.2500</td>\n      <td>0.0006</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>TF-IDF + Naive Bayes</td>\n      <td>0.8852</td>\n      <td>0.8852</td>\n      <td>0.8852</td>\n      <td>0.8890</td>\n      <td>0.8793</td>\n      <td>0.9524</td>\n      <td>0.3213</td>\n      <td>0.0945</td>\n      <td>0.0187</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>TF-IDF + Logistic Regression</td>\n      <td>0.9098</td>\n      <td>0.9098</td>\n      <td>0.9098</td>\n      <td>0.9204</td>\n      <td>0.8965</td>\n      <td>0.9700</td>\n      <td>0.2398</td>\n      <td>0.0696</td>\n      <td>0.0131</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                          model  ...  prediction_seconds_without_preprocessing\n0             Majority baseline  ...                                    0.0006\n1          TF-IDF + Naive Bayes  ...                                    0.0187\n2  TF-IDF + Logistic Regression  ...                                    0.0131\n\n[3 rows x 10 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Build the final comparison table\ntest_results = pd.DataFrame([\n    metric_summary(name, y_test, score) for name, score in test_scores.items()\n])\ntest_results['prediction_seconds_without_preprocessing'] = test_results['model'].map(prediction_seconds)\ndisplay(test_results.round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-068-interpret-the-selected-result-guide",
   "metadata": {},
   "source": "### 68. Interpret the selected result\n\nGenerate the narrative from the recorded predictions rather than typing scores from memory.\n"
  },
  {
   "cell_type": "code",
   "id": "step-068-interpret-the-selected-result",
   "metadata": {
    "portfolio_step": 68,
    "purpose": "Interpret the selected result",
    "execution_scope": "core",
    "execution_seconds": 0.0016
   },
   "execution_count": 68,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Validation selected TF-IDF + Logistic Regression.\nOn 9,916 held-out reviews: accuracy 90.98%; macro F1 0.9098.\nAt the default 0.5 rule it caught 89.65% of negative reviews; negative precision was 92.04%.\nThese are local held-out results, not official benchmark or deployment scores.\n"
    }
   ],
   "source": "# Interpret the selected result\nselected_row = test_results.set_index('model').loc[selected_name]\nselected_score = test_scores[selected_name]\nselected_prediction = (selected_score >= 0.5).astype(int)\nprint(f'Validation selected {selected_name}.')\nprint(f'On {len(y_test):,} held-out reviews: accuracy {selected_row.accuracy:.2%}; '\n      f'macro F1 {selected_row.macro_f1:.4f}.')\nprint(f'At the default 0.5 rule it caught {selected_row.negative_recall:.2%} of negative '\n      f'reviews; negative precision was {selected_row.negative_precision:.2%}.')\nprint('These are local held-out results, not official benchmark or deployment scores.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-069-plot-test-macro-f1-guide",
   "metadata": {},
   "source": "### 69. Plot test macro F1\n\nShow the complete executed comparison, including the majority baseline.\n"
  },
  {
   "cell_type": "code",
   "id": "step-069-plot-test-macro-f1",
   "metadata": {
    "portfolio_step": 69,
    "purpose": "Plot test macro F1",
    "execution_scope": "core",
    "execution_seconds": 0.2155
   },
   "execution_count": 69,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Plot test macro F1\nfig, ax = plt.subplots(figsize=(8, 4))\nchart_data = test_results.set_index('model')['macro_f1'].sort_values()\nax.barh(chart_data.index, chart_data.values,\n        color=[SCARLET if name == selected_name else NAVY for name in chart_data.index])\nax.set(xlim=(0, 1), xlabel='Macro F1 (0 to 1)', title='Held-out comparison at the fixed 0.5 rule')\nax.bar_label(ax.containers[0], fmt='%.3f', padding=3)\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-070-show-the-selected-classification-report-guide",
   "metadata": {},
   "source": "### 70. Show the selected classification report\n\nBoth class rows matter. The support column gives the number of true examples in each class.\n"
  },
  {
   "cell_type": "code",
   "id": "step-070-show-the-selected-classification-report",
   "metadata": {
    "portfolio_step": 70,
    "purpose": "Show the selected classification report",
    "execution_scope": "core",
    "execution_seconds": 0.0208
   },
   "execution_count": 70,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>precision</th>\n      <th>recall</th>\n      <th>f1-score</th>\n      <th>support</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>negative</th>\n      <td>0.9204</td>\n      <td>0.8965</td>\n      <td>0.9083</td>\n      <td>4939.0000</td>\n    </tr>\n    <tr>\n      <th>positive</th>\n      <td>0.8999</td>\n      <td>0.9230</td>\n      <td>0.9113</td>\n      <td>4977.0000</td>\n    </tr>\n    <tr>\n      <th>accuracy</th>\n      <td>0.9098</td>\n      <td>0.9098</td>\n      <td>0.9098</td>\n      <td>0.9098</td>\n    </tr>\n    <tr>\n      <th>macro avg</th>\n      <td>0.9101</td>\n      <td>0.9098</td>\n      <td>0.9098</td>\n      <td>9916.0000</td>\n    </tr>\n    <tr>\n      <th>weighted avg</th>\n      <td>0.9101</td>\n      <td>0.9098</td>\n      <td>0.9098</td>\n      <td>9916.0000</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "              precision  recall  f1-score    support\nnegative         0.9204  0.8965    0.9083  4939.0000\npositive         0.8999  0.9230    0.9113  4977.0000\naccuracy         0.9098  0.9098    0.9098     0.9098\nmacro avg        0.9101  0.9098    0.9098  9916.0000\nweighted avg     0.9101  0.9098    0.9098  9916.0000"
     },
     "metadata": {}
    }
   ],
   "source": "# Show the selected classification report\nclass_report = pd.DataFrame(classification_report(\n    y_test, selected_prediction, labels=[0, 1], target_names=['negative', 'positive'],\n    output_dict=True, zero_division=0,\n)).T\ndisplay(class_report.round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-071-count-the-selected-model-s-errors-guide",
   "metadata": {},
   "source": "### 71. Count the selected model’s errors\n\nRows are actual sentiment; columns are predicted sentiment. This orientation stays consistent in the chart.\n"
  },
  {
   "cell_type": "code",
   "id": "step-071-count-the-selected-model-s-errors",
   "metadata": {
    "portfolio_step": 71,
    "purpose": "Count the selected model’s errors",
    "execution_scope": "core",
    "execution_seconds": 0.0075
   },
   "execution_count": 71,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>predicted negative</th>\n      <th>predicted positive</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>actual negative</th>\n      <td>4428</td>\n      <td>511</td>\n    </tr>\n    <tr>\n      <th>actual positive</th>\n      <td>383</td>\n      <td>4594</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                 predicted negative  predicted positive\nactual negative                4428                 511\nactual positive                 383                4594"
     },
     "metadata": {}
    }
   ],
   "source": "# Count the selected model’s errors\nselected_cm = confusion_matrix(y_test, selected_prediction, labels=[0, 1])\ndisplay(pd.DataFrame(selected_cm, index=['actual negative', 'actual positive'],\n                     columns=['predicted negative', 'predicted positive']))\nassert int(selected_cm.sum()) == len(y_test)\nassert np.isclose(np.trace(selected_cm) / selected_cm.sum(), selected_row.accuracy)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-072-plot-the-confusion-matrix-guide",
   "metadata": {},
   "source": "### 72. Plot the confusion matrix\n\nA missed negative is the top-right cell. A false negative-feedback alert is the bottom-left cell.\n"
  },
  {
   "cell_type": "code",
   "id": "step-072-plot-the-confusion-matrix",
   "metadata": {
    "portfolio_step": 72,
    "purpose": "Plot the confusion matrix",
    "execution_scope": "core",
    "execution_seconds": 0.1184
   },
   "execution_count": 72,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Plot the confusion matrix\nfig, ax = plt.subplots(figsize=(5, 4))\nax.imshow(selected_cm, cmap='Blues')\nfor (row, column), count in np.ndenumerate(selected_cm):\n    ax.text(column, row, f'{count:,}', ha='center', va='center',\n            color='white' if count > selected_cm.max() / 2 else NAVY, fontsize=13)\nax.set(xticks=[0, 1], yticks=[0, 1], xticklabels=['Negative', 'Positive'],\n       yticklabels=['Negative', 'Positive'], xlabel='Predicted sentiment',\n       ylabel='Actual sentiment', title='Selected model: held-out errors')\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-073-explain-the-error-counts-in-words-guide",
   "metadata": {},
   "source": "### 73. Explain the error counts in words\n\nConnect the matrix to a negative-feedback queue without implying these are actual customer complaints.\n"
  },
  {
   "cell_type": "code",
   "id": "step-073-explain-the-error-counts-in-words",
   "metadata": {
    "portfolio_step": 73,
    "purpose": "Explain the error counts in words",
    "execution_scope": "core",
    "execution_seconds": 0.0003
   },
   "execution_count": 73,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Default rule: 511 negative reviews were classified as positive.\n383 positive reviews would be incorrectly flagged as negative.\nThe cost of these two mistakes can differ by organisation and use case.\n"
    }
   ],
   "source": "# Explain the error counts in words\ntrue_negative, missed_negative = selected_cm[0]\nfalse_alert, true_positive = selected_cm[1]\nprint(f'Default rule: {missed_negative:,} negative reviews were classified as positive.')\nprint(f'{false_alert:,} positive reviews would be incorrectly flagged as negative.')\nprint('The cost of these two mistakes can differ by organisation and use case.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-074-plot-negative-class-ranking-quality-guide",
   "metadata": {},
   "source": "### 74. Plot negative-class ranking quality\n\nUse negative scores and negative labels so these curves directly describe the review queue.\n"
  },
  {
   "cell_type": "code",
   "id": "step-074-plot-negative-class-ranking-quality",
   "metadata": {
    "portfolio_step": 74,
    "purpose": "Plot negative-class ranking quality",
    "execution_scope": "core",
    "execution_seconds": 0.4375
   },
   "execution_count": 74,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Plot negative-class ranking quality\nnegative_truth = (y_test == 0).astype(int)\nfig, axes = plt.subplots(1, 2, figsize=(10, 4))\nRocCurveDisplay.from_predictions(negative_truth, 1 - selected_score, ax=axes[0], curve_kwargs={'color': NAVY})\nPrecisionRecallDisplay.from_predictions(negative_truth, 1 - selected_score, ax=axes[1], color=SCARLET)\naxes[0].set_title('Negative-review ROC curve')\naxes[1].set_title('Negative-review precision–recall')\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-075-inspect-probability-calibration-guide",
   "metadata": {},
   "source": "### 75. Inspect probability calibration\n\nA high score is not automatically a trustworthy probability. This diagnostic assesses the frozen model; it does not recalibrate on test data.\n"
  },
  {
   "cell_type": "code",
   "id": "step-075-inspect-probability-calibration",
   "metadata": {
    "portfolio_step": 75,
    "purpose": "Inspect probability calibration",
    "execution_scope": "core",
    "execution_seconds": 0.1975
   },
   "execution_count": 75,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Inspect probability calibration\nfraction_positive, mean_score = calibration_curve(y_test, selected_score, n_bins=10, strategy='quantile')\nfig, ax = plt.subplots(figsize=(5, 4))\nax.plot([0, 1], [0, 1], '--', color=SLATE, label='Perfect calibration')\nax.plot(mean_score, fraction_positive, 'o-', color=NAVY, label=selected_name)\nax.set(xlabel='Mean predicted positive score', ylabel='Observed positive fraction',\n       title='Test calibration diagnostic', xlim=(0, 1), ylim=(0, 1))\nax.legend(fontsize=8)\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-076-define-a-bootstrap-interval-for-a-fixed-model-guide",
   "metadata": {},
   "source": "### 76. Define a bootstrap interval for a fixed model\n\nResample test rows 1,000 times. The interval measures row-sampling uncertainty conditional on this trained model, not training-seed or domain-shift uncertainty.\n"
  },
  {
   "cell_type": "code",
   "id": "step-076-define-a-bootstrap-interval-for-a-fixed-model",
   "metadata": {
    "portfolio_step": 76,
    "purpose": "Define a bootstrap interval for a fixed model",
    "execution_scope": "core",
    "execution_seconds": 0.0004
   },
   "execution_count": 76,
   "outputs": [],
   "source": "# Define a bootstrap interval for a fixed model\ndef bootstrap_mean_interval(values, seed=RANDOM_STATE):\n    values = np.asarray(values)\n    rng = np.random.default_rng(seed)\n    samples = rng.choice(values, size=(BOOTSTRAP_REPEATS, len(values)), replace=True)\n    return np.quantile(samples.mean(axis=1), [0.025, 0.975])\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-077-estimate-test-accuracy-intervals-guide",
   "metadata": {},
   "source": "### 77. Estimate test accuracy intervals\n\nReview-level independence is an assumption. Hidden shared films/authors could make these intervals too narrow.\n"
  },
  {
   "cell_type": "code",
   "id": "step-077-estimate-test-accuracy-intervals",
   "metadata": {
    "portfolio_step": 77,
    "purpose": "Estimate test accuracy intervals",
    "execution_scope": "core",
    "execution_seconds": 0.3086
   },
   "execution_count": 77,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>model</th>\n      <th>accuracy</th>\n      <th>bootstrap_95_lower</th>\n      <th>bootstrap_95_upper</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>Majority baseline</td>\n      <td>0.5019</td>\n      <td>0.4916</td>\n      <td>0.5118</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>TF-IDF + Naive Bayes</td>\n      <td>0.8852</td>\n      <td>0.8790</td>\n      <td>0.8918</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>TF-IDF + Logistic Regression</td>\n      <td>0.9098</td>\n      <td>0.9044</td>\n      <td>0.9156</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                          model  ...  bootstrap_95_upper\n0             Majority baseline  ...              0.5118\n1          TF-IDF + Naive Bayes  ...              0.8918\n2  TF-IDF + Logistic Regression  ...              0.9156\n\n[3 rows x 4 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Estimate test accuracy intervals\ninterval_rows = []\nfor name, score in test_scores.items():\n    correct = (score >= 0.5) == y_test\n    lower, upper = bootstrap_mean_interval(correct)\n    interval_rows.append({'model': name, 'accuracy': correct.mean(),\n                          'bootstrap_95_lower': lower, 'bootstrap_95_upper': upper})\naccuracy_intervals = pd.DataFrame(interval_rows)\ndisplay(accuracy_intervals.round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-078-compare-accuracy-with-a-paired-bootstrap-guide",
   "metadata": {},
   "source": "### 78. Compare accuracy with a paired bootstrap\n\nUse the same rows for both models. A descriptive interval that excludes zero supports a difference on this holdout, not universal superiority.\n"
  },
  {
   "cell_type": "code",
   "id": "step-078-compare-accuracy-with-a-paired-bootstrap",
   "metadata": {
    "portfolio_step": 78,
    "purpose": "Compare accuracy with a paired bootstrap",
    "execution_scope": "core",
    "execution_seconds": 0.0919
   },
   "execution_count": 78,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Accuracy difference versus TF-IDF + Naive Bayes: 2.46%.\nPaired bootstrap 95% interval: 1.91% to 3.04%.\nThe selected model remains the validation choice regardless of this test comparison.\n"
    }
   ],
   "source": "# Compare accuracy with a paired bootstrap\ncomparison_name = ('TF-IDF + Naive Bayes' if selected_name != 'TF-IDF + Naive Bayes'\n                   else 'TF-IDF + Logistic Regression')\nselected_correct = ((selected_score >= 0.5) == y_test).astype('int8')\ncomparison_correct = ((test_scores[comparison_name] >= 0.5) == y_test).astype('int8')\naccuracy_difference = selected_correct - comparison_correct\ndifference_interval = bootstrap_mean_interval(accuracy_difference)\nprint(f'Accuracy difference versus {comparison_name}: {accuracy_difference.mean():.2%}.')\nprint(f'Paired bootstrap 95% interval: {difference_interval[0]:.2%} to {difference_interval[1]:.2%}.')\nprint('The selected model remains the validation choice regardless of this test comparison.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-079-assess-the-locked-triage-rule-on-the-test-set-guide",
   "metadata": {},
   "source": "### 79. Assess the locked triage rule on the test set\n\nApply the validation-selected threshold once. Do not choose a better-looking threshold using these test results.\n"
  },
  {
   "cell_type": "code",
   "id": "step-079-assess-the-locked-triage-rule-on-the-test-set",
   "metadata": {
    "portfolio_step": 79,
    "purpose": "Assess the locked triage rule on the test set",
    "execution_scope": "core",
    "execution_seconds": 0.0138
   },
   "execution_count": 79,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>policy</th>\n      <th>threshold</th>\n      <th>flagged</th>\n      <th>negative_recall</th>\n      <th>negative_precision</th>\n      <th>missed_negatives</th>\n      <th>false_alerts</th>\n      <th>relative_cost_units</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>Reference negative threshold 0.50</td>\n      <td>0.50</td>\n      <td>4811</td>\n      <td>0.8965</td>\n      <td>0.9204</td>\n      <td>511</td>\n      <td>383</td>\n      <td>2938</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>Validation-locked illustrative triage</td>\n      <td>0.25</td>\n      <td>5906</td>\n      <td>0.9698</td>\n      <td>0.8110</td>\n      <td>149</td>\n      <td>1116</td>\n      <td>1861</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                                  policy  ...  relative_cost_units\n0      Reference negative threshold 0.50  ...                 2938\n1  Validation-locked illustrative triage  ...                 1861\n\n[2 rows x 8 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Assess the locked triage rule on the test set\ntest_triage = pd.DataFrame([\n    dict(policy='Reference negative threshold 0.50',\n         **triage_summary(y_test, selected_score, 0.50)),\n    dict(policy='Validation-locked illustrative triage',\n         **triage_summary(y_test, selected_score, triage_threshold)),\n])\ndisplay(test_triage.round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-080-interpret-the-triage-trade-off-guide",
   "metadata": {},
   "source": "### 80. Interpret the triage trade-off\n\nQueue volume is an operational constraint; the cost comparison is based on stated assumptions.\n"
  },
  {
   "cell_type": "code",
   "id": "step-080-interpret-the-triage-trade-off",
   "metadata": {
    "portfolio_step": 80,
    "purpose": "Interpret the triage trade-off",
    "execution_scope": "core",
    "execution_seconds": 0.0015
   },
   "execution_count": 80,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "At threshold 0.25: 5,906 reviews are flagged (59.6% of this test set).\nMissed negatives: 511 → 149; false alerts: 383 → 1,116.\nAssumed cost units: 2,938 → 1,861.\nThis is a movie-review simulation; no real service-time or revenue benefit was measured.\n"
    }
   ],
   "source": "# Interpret the triage trade-off\nreference, triage = test_triage.iloc[0], test_triage.iloc[1]\nprint(f'At threshold {triage_threshold:.2f}: {int(triage.flagged):,} reviews are flagged '\n      f'({triage.flagged / len(y_test):.1%} of this test set).')\nprint(f'Missed negatives: {int(reference.missed_negatives):,} → {int(triage.missed_negatives):,}; '\n      f'false alerts: {int(reference.false_alerts):,} → {int(triage.false_alerts):,}.')\nprint(f'Assumed cost units: {int(reference.relative_cost_units):,} → {int(triage.relative_cost_units):,}.')\nprint('This is a movie-review simulation; no real service-time or revenue benefit was measured.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-172",
   "metadata": {},
   "source": "## Explain and inspect errors\n\nModel explanations show learned associations. They do not show that a word causes a\nperson to feel positively, and they do not establish whether a review is true. Inspecting\nfailures is more informative than presenting only a favourable example.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-081-inspect-logistic-regression-feature-weights-guide",
   "metadata": {},
   "source": "### 81. Inspect logistic-regression feature weights\n\nPositive coefficients support the positive class; negative coefficients support the negative class. Correlated phrases share evidence.\n"
  },
  {
   "cell_type": "code",
   "id": "step-081-inspect-logistic-regression-feature-weights",
   "metadata": {
    "portfolio_step": 81,
    "purpose": "Inspect logistic-regression feature weights",
    "execution_scope": "core",
    "execution_seconds": 0.0131
   },
   "execution_count": 81,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>negative_feature</th>\n      <th>negative_weight</th>\n      <th>positive_feature</th>\n      <th>positive_weight</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>worst</td>\n      <td>-11.044</td>\n      <td>great</td>\n      <td>10.940</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>bad</td>\n      <td>-10.931</td>\n      <td>excellent</td>\n      <td>8.716</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>awful</td>\n      <td>-10.338</td>\n      <td>perfect</td>\n      <td>7.951</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>boring</td>\n      <td>-9.613</td>\n      <td>wonderful</td>\n      <td>6.820</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>the worst</td>\n      <td>-8.719</td>\n      <td>amazing</td>\n      <td>6.636</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>poor</td>\n      <td>-8.385</td>\n      <td>best</td>\n      <td>5.893</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>waste</td>\n      <td>-7.628</td>\n      <td>the best</td>\n      <td>5.424</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>terrible</td>\n      <td>-7.582</td>\n      <td>loved</td>\n      <td>5.384</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>worse</td>\n      <td>-6.900</td>\n      <td>enjoyable</td>\n      <td>5.296</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>disappointment</td>\n      <td>-6.534</td>\n      <td>fun</td>\n      <td>5.106</td>\n    </tr>\n    <tr>\n      <th>10</th>\n      <td>stupid</td>\n      <td>-6.454</td>\n      <td>brilliant</td>\n      <td>4.932</td>\n    </tr>\n    <tr>\n      <th>11</th>\n      <td>dull</td>\n      <td>-6.451</td>\n      <td>hilarious</td>\n      <td>4.853</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "   negative_feature  negative_weight positive_feature  positive_weight\n0             worst          -11.044            great           10.940\n1               bad          -10.931        excellent            8.716\n2             awful          -10.338          perfect            7.951\n3            boring           -9.613        wonderful            6.820\n4         the worst           -8.719          amazing            6.636\n5              poor           -8.385             best            5.893\n6             waste           -7.628         the best            5.424\n7          terrible           -7.582            loved            5.384\n8             worse           -6.900        enjoyable            5.296\n9    disappointment           -6.534              fun            5.106\n10           stupid           -6.454        brilliant            4.932\n11             dull           -6.451        hilarious            4.853"
     },
     "metadata": {}
    }
   ],
   "source": "# Inspect logistic-regression feature weights\ncoefficients = logistic_model.coef_.ravel()\npositive_indices = np.argsort(coefficients)[-12:][::-1]\nnegative_indices = np.argsort(coefficients)[:12]\nfeature_weights = pd.DataFrame({\n    'negative_feature': feature_names[negative_indices],\n    'negative_weight': coefficients[negative_indices],\n    'positive_feature': feature_names[positive_indices],\n    'positive_weight': coefficients[positive_indices],\n})\ndisplay(feature_weights.round(3))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-082-plot-interpretable-term-associations-guide",
   "metadata": {},
   "source": "### 82. Plot interpretable term associations\n\nThese weights belong to the logistic baseline even if an optional neural model wins validation.\n"
  },
  {
   "cell_type": "code",
   "id": "step-082-plot-interpretable-term-associations",
   "metadata": {
    "portfolio_step": 82,
    "purpose": "Plot interpretable term associations",
    "execution_scope": "core",
    "execution_seconds": 0.2566
   },
   "execution_count": 82,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABKEAAALaCAYAAADp8kAfAAAAOnRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjEwLjgsIGh0dHBzOi8vbWF0cGxvdGxpYi5vcmcvwVt1zgAAAAlwSFlzAAAXEgAAFxIBZ5/SUgAA0QhJREFUeJzs3XlYVGX/x/HPIKviigIuJLgrLrjjCpoo5RKp5VqpWdZjqamVLZbVU6YtmlqWZuGjqZk75lYqauIuqOFOgrjv4sr++8Mfk+MMCsgwUO/XdXFdzn3Ouc93DscZ5jP3uY8hPT09XQAAAAAAAIAV2dm6AAAAAAAAAPzzEUIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6QigAAAAAAABYHSEUAAAAAAAArI4QCgAAAAAAAFZHCAUAAAAAAACrI4QCAAAAAACA1RFCAQAAAAAAwOoIoQAAAAAAAGB1hFAAAAAAAACwOkIoAAAAAAAAWB0hFAAAAAAAAKzO3tYFAACAf5cnX3xHEbuijY97dGqjSR8MtWFF0vFTZ9W48yCTtkXffaQWjerYqCLbatTpBcWfPm98PPLFHnp9UC8bVpR1Q97/Sj8vX2983LyhrxZP+9iGFT1YamqqZi1ao8WrN+nQX/G6eu2G0tLSjMvP7lryrzpHN+/cp66DRpu07Qj7To+U88hWPwXpPM5vv9/8+Dpd0OXFMc1v5xFgCSEUgDy37+BfmrP0d23fc0Dxp87pxs3bci3iohLFXFWqRDFVq+Ql36reqlezsprWr2W2/ead+xSx60/j42JFi2hQ7y55+RT+kf489JdWhm8zacuvf6wDWTFv2VrFnz5nfOxbzUePt/G3YUXIieOnzurnsHUmbS/27qziRV1tVFHue2HUZ/p13daH7ufqteuaNifMpK1H57aZhjfZXR8FF+/xAPILQigAeeqDiaGaOnup0tPTTdqvJFzXlYTrij1xRrv/PCxJKlW8qA6sm2XWR8SuP/X5tJ+Nj73KliGEygV/Hjpmclwl/kBFwfbz8nVm3zoTQhU88afOmb029ejc9h8TQm2L3J8rAZQkXb12w+xYNW9Y+z4hVPbWR8HFezyA/IIQCkCe+fanZfpm1hJblwHAxr77ZKRuJyUZHxdxcbFhNXeUcy+tHWHfmbS5u5W0UTW2t2zGWKWkphofFy9axIbV/LNlfPGSwb5QIU37dKR8q3rLrtDf07dyjv6z5bffb358nQbwz0AIBSBPpKWladKPC03afKt565XnuqqaTwUVdnHWlYTrOhp7Utui9mtdxG7dvp2USW8ACjL30vnvg7O9fSFGf9ylnEdpW5fwr3Hj1m2Tx55lSqlj22Zm63GO/rPlt99vfnydBvDPQAgFIE8ciT2pi5evmrTN/PJteZV1N2lrULuanu7URmlpadq+54DJsnsnm80Qf/q8PBqGmLR99f6r6tnlUUmWJyYd0r+7ps0J05I1m3Qs/rRu3LxtNnFjcnKKlv72h1as36o9+2N08fJVpUsqVaKoalevpOCAJur+eKCcHB3MaspsUlUnBwdNCl2k3zbt0Jnzl1TExVmN6tXQsAFPqWGdapkev6OxJzVhxnxt2r5XVxKuq0ypEmrTvL5ee/4ppaalPdQklJ99N9dsiH6Ge4+rpUldr12/qTnLfte6zbu1/0icriRck6Ojg8p7llGLhrU14OnHVdWngsX+LU3SOfH9VzV78W/6efk6Hf4rXgnXbxp/n/OWrdXQDyab9HF21xJtjzqgyTMXaeeeg7qdlKxqPhXU76nH1Ov/zwFJ+uXX9QpdsEoHY45LkmpV8dbzPTsqpH3LLB2ne924eUvLfo/Q3oMx2n84VmcvXNaVhOu6dv2mXJyd5F66hGpXr6SQoJZ6rE1TGQwGi/0cjT2p0AUrtXX3fh0/dVY3bt6Wi7OT3EoWUzkPN9WrWUX1fauqXatGKuLibLLtxu179HPYOkVFH9Xp8xeVmJisYq6F5VayuCpW8FR93ypqWLu62jSvb3HfZ85f0qxFq7Vx217FxJ1UwvWbcnZyVDkPNzWuV0M9uzyqxnVr3Pc4pKSkavm6Lf///+SoLly+quTkFLmVLCavsu5q1aSunurYRt4VPI3bZGVy1hXrtipy/xHtO/iXTpw+r8sJ13Q14Ybs7QupVPGiqlHlEQW1bKSnO7VRkcKm39Df+38+w8/L15u9hmRMdpydyVzzy2vDw5w7D/KgCZ1z+zXOEkv/3zPc+7vKygS/2yL369uflmn7ngNKuHZDZd3d9FhgUw1/4en7Xtr3MK9x95PZe9qJM6bvaRnP7X7nqKXfR4Z725s39NXIF3tma/17J3ZPTU3V8rVbtHxthCL3H9WFS1eVnp6uMm4l1KhOdfXq8qgC/P0ye+qSpB17D2pK6CLt2HNQN27dVgXPMurUrrlefa7rfbezlohdf2r+r+u1c88hnTl/SbcTk1SsaBFV86mg1k3r6dluHVS6ZPFMt7+dmKRvf1qqRas2Ke7EGbk4O6lujUp6vmdHdWjd5IGve1l9DYraf1SzF6/Rjj0HdeLMed26naiiRQqrVIli8ipbRn6+VeVXq4rat2ose/tCOX6Pz84k2rEnzuinJb9py+5o/XX8tBKu3VCRws4qU6qE6tWqrLbNG6jbYwGZHjtL7q3bq2wZ7Vw+Xas3bte0OWHaeyBG6el3vtB8qe8TeiywqaQ75+aPv6zU3GVrFRN3So4O9qpXq4peea6rAprWu+8+H/YcuHU7UVNnLdGSNX8o7uRZFXFxVn3fqnqp7xNq1aRutp7/gaNxmr34N22NjNaJ0+d14+ZtFStaRNUrealDQBM907V9tl/XgfyCEApAnki4dsOs7fqNW5mub2dnJ//6vlap5dqNW+o8YJT2HIjJdJ2DMcf1wpvjdfjYCbNlp85e1KmzF7Vm4w5N/GGBpn/6uvxqVXngfjdt36v3v/xR127cNLYlJiVrzcYdCt8SqVkT3lFgM/OwYM3GHRr45nglJiUb206cOa9Zi9Zoyeo/9OmoQWbb5JU1G3do6JhJunT1mkl7UnKKDsUc16GY4wpdsEqvv9hDw1/o8cD+klNS9cywj/X75l1ZruGH+Sv0zmffm9xJKmr/UQ37YLJ27TukcaMG6eV3J2jpmj9Mttu+54C27zmgv46f0vCBT2d5fxmOxZ/WsEw+IF+7cVPXbtxUTNwpLV3zh5o3rK1ZE96RaxHTsOTXdVv00ttfKCk5xeL2sSfOGD8E/Bo6To3qVDeuM2bCj5o6e6nZvi9dvaZLV6/pSOwJ/f7HTjk5Ouj4ll/M1vtx/gq9P+FHk/NKkpJTUnTor5s69Fe8Zi/+TU92aKUv3v2PWdAj3bnJwMvvfKkjsZn/P9kWdSBHNw8Y+sEkJVy/adaenJKik7cTdfLsBa3dvFtf/2+xZk18VzWrVMxW/zmVX14bHubcsZacvsblhfHfztWX3883mY8w7uRZffvTMq3fGqWVoeMsnuO5/Rr3T3A09qReGPWZ9h+JNVsWf+qc4k+d0+LVm/R4G39N/mCo2euedOd1++3x001+H0fjTmrijF+0eNVGjXyxpzWfgokrCdc1ZMwkrd6w3WzZxctXteXyVW3ZHa3JoYs0btQgPd2pjYX1EvTUf95T9OFYY9vtxCRt2LZHG7bt0Ut9cmfeyu/nLde7n8/IdF7Nv46f0oZteyRJ+1b/aPXRTCkpqfp06hx9/b/FJu/Bd9d0JPaE1mzcke0QypKPJs3UlJmLTdq2Ru7X1sj9entwX73Qq7N6D/lQW3b/HZ7dup2ojdv2aNP2vZo0ZojF319unAPnL11R95feM37RJd05B37fvEtrI3brjZeyNv9WUnKy3p/wo374eYXFWiJ2XVXErj/17eylmjH+zWyH+0B+YPfgVQDg4ZVxK2HW1m3QaH34Vah+/2Onzl+6kme1zJj3630DqLiTZ9V10GiLHzLvdfzkWT318ns6fCz+geuO+O83Jh/O7paUnKKRH3+j1LvmYJGkmLiTenHUZ2ZBQYZrN25mGoRY29rNu9Rv5FizD2f3SktL07hv5+rL6Za/ib3bst82ZyuAkqS3x083++M3w6xFa/T04DFmAdTdPp82T8fiT2drn9kVsetPvfnptyZtScnJGvHfb8xChKyI2n/UYgCVVT/MX6FR46Zlel7dbfHqTRr45mdm5+b+I7HqOuhdiwFUXoo/fV7PDf9EyTk4jtmVX14bHubcsaacvMblhR17DumL6T+bfXDPcCjmuCaHLjJrt8ZrXEF34vR5PfniuxYDqHutWL9Vz78xzux3/seOfWYB1N3iTp7VG2O/tbgstyUmJeuZYR9bDB/udfPWbb36/ldasGKDSXt6erpeeucLkwDqXt/+tEw79hx6qFrPXbisMRNCMz1utjDi4280OXRhpu/BuenEmQtmAdTdxn07xyyAult6erreGj9N1+75ciO3zoFBb31hEkDdu3zc1DmKjD76wH0MHTPZYgB1r9PnLurpwe/r0F8Pfo8B8htGQgHIE94VPFWzSkUdOBpnbLt4JUFf/2+Jvv7fEklSOQ83NapbQ+1bNdLjbZuZDTN+f1h/jRzUU9PmhGn63OXG9rLublo24xOTdd1KFMu0lpTUVBUtUlgjB/VUQNN6SklJ1e4/DxsnAH17/HSzSwef6dpeTz0eKHt7ey1Zs8nkltYJ12/qzbHfml2ucK/09HR1DW6tQX26KDk5RZ98PctkqHv86fPasfegyQiwsV//pFuJpnNj9XriUQ146nHZ2Rk0a9EahS5Ydd/9PsiLvTurR+e2Wv77Fn3wVajJsnsnSc2YnPjW7US99uEUpab+/Ydng9rV9Gq/rqrqXUFXr9/QnCW/66clvxmXfz79Zz3RvqUqVyyfaS0pqalysLfX0AHdFBzQVIUK2Sn68DF5Vyib6TaF7Oz07qvP6NGWDbXv4F8a/tHXun3XMftjxz4VKeysj19/QQ3rVNPaP3ZpzMS/n2dqapoWr96U7dFQBoNBvtW8FdSqsfxqVVHpksVVqkRRJSYlK/70Oc1e/JvWbNxhXH/Rqk1699VnVdbdTZJ08OhxXb7rw23J4kX18esDVbt6JdkXstOFS1d16Fi8tkXu19rNu032vfWeP7Lr1qikd155Rl7l3JWSmqYz5y4q+nCsNu/cZ3ZZ66mzF/TBXc9fuvN7fXfIs2pYp7pOn72g8d/ONQlq10Xs1i+/hhsvcU1PT9drH04xG6nUsnEdvdi7s6pULK/bScnas/+o5i1bm63jmqG8Zxk9Wa+mmjeqLXe3EnIrWVwGg3T+4hWt3rhD0+cuN37wiTt5VsvXbdGTHVpJ+ntS7UFvfWEy6XOnR5vp/WH9TPZTzj3rcx/ll9eGhzl3rCknr3H307ldczVvVFu79h3WS29/YbJs2fefqKyHm/Hx/SZNTk5JUWEXZ7035Fm1aFxH+w7+pTfHfmcSmC1atVGj/tPH+Nhar3F3y+p7WlYmhG5Yp7p2hH2n02cvqsvAt02WffvJCJPREs6OjipWtEi21s/wzmfTde7iZeNjj9Il9fqgXmpQp5qUnq4NW/fo06k/GQPu8K1R+nn5evV+op1xmw8m/mgWpAzq00VPdmilm7du67s5YVkKBHLDjHm/mr1G+tevpWEDnpJ7mZLaujta/508Szfvmrfr7fHTFNSqofESzt//2KWN/z/6KENV7woa81o/eVcoq137Dun9CT+a/J/NiR17Dyo55e/g2aucuz4a8byq/P/5du7iZR04Gqctu6O1fkukcb2cvMdnxZqNO8xe312LuOjVfl3Vpll9FS1SWCfOnNe6zbu1fG1ElvvNTHp6uooUdtZ/Rw5Uo7rVtWn7Xr09frpxeWpqmrbsjpa7W0mNffMFVfGuoHnL1pp8YXP9xi2t3rhd3R8PNLblxjmwZtMObd65z6SPiuU9NGZYf3l7eWrnvkP6YGLofa8AkO4Et4tWbTQ+NhgMer7HnWkDShYvqsPH4jX+27nGv6Wv37ilNz6ZqqXff5JZl0C+RAgFIM98/u5/9NTL75u8kd/t1NmLWvbbZi37bbPcvvxRH44YYPKHglvJYnIrWczsjyT7QnbZnsxz6sfDFdSqkfFxnRqVJN35lvf3P3aarNv98QB9/s5/jI8b1qmm24lJ+t/C1ca2iF3ROnwsXtV8vDLdZ6O61fXNf18zzg00fdwb8m33nMk6+4/EGT+g3bh5SyvDt5ks969fSxPfe9X4eNxbL+n4qXNaF5HzD5rFi7qqeFFXlSpR1GxZZsf113VbdPbC3x9E3EoW18JvP1JhFydjW6M61XU09oS2Rd354y41NU2zFq3RmNf637eeD0cM0ICnHzc+rlXV+77rD+jxuF5+JkSSVM3HS7+u22J2u/NRL/cxzg9VzcdLC1Zu0J+HjhmX77/PN9iZ8a3mo3VzJ1pcVquqt9q1aKgabZ8xBjVpaWnaFnXAOAdVyj2jA+r7VjW5XKFyxfJqWr+Wnu3aQcnJKUq965vme7ftEtTS5DKn6pW8FODvp/88G2L2R++8ZWtNQjpJ+n78G2rd5M5cGb5VvdWsga8adx6ki1cSjOuELlhlDKF27j2kqP2m3+h2erSZvh/3hsncV75VvdX7iXa6eu26xeN0P+E/f2WxvZqPl1o0qqO/jp/Sb5v+/r+6ZVe0MYTKmFTb2cl0TqYiLs45nvg3P702PMy5Y03ZfR4PUqSwi4oUdlH8qXNmy8p6uGXrd/nBa/31bLcOku6cQ2fPXzb5QB538qxu3Lpt/PLDmq9xf/eZe+9pzk6OmW7j7lbC4rLsrn/q7AWtuiccmvnl26rvW9X42LeajxKu39CEGX9fAhz6y0pjCHX4WLz2HvzLpI9eTzyqD4cPMD5u3rC2Hu392n1HFuWW0AUrTR4/Ut5DC6Z+KAeHOx+RfKt6q1SJYiYh6NVrN7R41Sb1e+oxSdIvK8JN+rCzs9P/JrytSo+UkyRV8S4vZydHvfjW5w9Va0qK6f/7ts0aGOdBkqSqPhXUolEdDezZSbduJ8rB/s5zyMl7fFZ8P2+5yWMHe3st/PYjk0uQKz1STq2b1NNrA5/K8X7uNurlPsZzqZqPl+YtW2t2Po1/+yXjcRk95FnNX77e5L0s+nCsuv/9J0aunAP3jowqVMhOv0z9UBXL3zm+tap6y61EMQ14fdx9n98P801HQPV/6jF9/PpA4+Mq3uVVq0pFNQ152di2NXK/DhyNy7NL0oHcQAgFIM80qlNdK2eO15gJPyp8a9R9h5RfvJKgwaMnytnJUZ0ebZ6rddSpXskkgLrb1kjzYdzPdG1v1vZs1w4mHzQlaevu/ff9oNn/qcdNPqCXLllcpYoXNbnU40rC3x/W9xyIMfuw2ffJILN++z4Z9FAhVE7cPbpBujNPgU/LB8+HsiVy/32Xl3ErYfE53s9TdwWVkuU/qu+dv6HSI+VMQqgrOQhJpDtznc0NW6vwLVE6EntCFy8n6NbtxEzP7dNnLxj/Xb2Sl1ycHI0j3dZF7FaX599Sg9rVVKliOVWpWF61q/moWNEicnCw191xSt2alU36Hf/tHO3Ye1C1qlRU5YrlVNXHS7WqVpSjg4PZfCwR94yi8q7gaQygMhQp7KIng1ubfMjYcyBGN28lqrCLk/645xtf6c6Hg8wmX7/fpM+ZSU1N1bLfI7QyfJsOHInV6XOXdPP2bZORKXc7de6Cxfbckp9eGx7m3LGm7D6PvFKksLN6dG5r0lbF23y00tWE68YQylqvcQXZvcdEkoKfff2B2+07dEw3bt5SkcIu2r3vsNny3l3amTw2GAzq2flRjf5iRs6LzYJTZy8o7uRZk7Zendsaw4cMTwS10KhPvzM5d7dE7jcGEHePtpTuTOSeEUBl6PRoMxUvWkRXLcyPmVV1a1aWwWAwvr/MWrxGcSfPqHb1Snde970ryLeajwq7OMnF2ekBvT2ctLQ0bb3nXH+8rX+mc+Dl5D3Akqc6Bpo8fqS8h0kIVaKYq9rf9fddoUKF9Eh5D5MQ6u7fY26dA5H3nAMtGtUxBlAZHgtsqpLFi2Y6Ii41NVXbo0xHZP0wf4VZMGXJ1t3RhFAoUAihAOSpGpUf0bwp7+v4qbNaHxGpbVEHtPvPw5nOyTP+u3m5HkLVruGT6bIz5y+ZtVm6FOzePy4y2/ZuVX3MP/Q4OztKd13dc/fcGecuXslSLXffdSyvnDl/MUfbnbtw/2NUvZKXHB2y95G5YgXT38W9f3wXcy2sEsVM/wB2cXI0eZyTeWp27TusZ1772OzyrPu5+1bsRQq7aNTgvnr/yx+MbduiDhhHVUh3vlFvULuqXuzVWU/cdRe/1k3qKTigiXFUQlJyilZv2G5yCYuzk6PaNKuvYc8/ZfLB4Ow956mlc9lSe1pams5fuqKK5T3M+nBxdsrR3cEyc+HyVfV+9cP7zt12rxs3LY+wzC356bXhYc4da8ru88grXmXdze5U6OzsaLbe3aG/tV7jCrKcHpO0tDSdv3hFRQq76Pwl89fLRyz8n7HUltss/b+saOH91M7OThXKljEJIO5+DTx/z3u1pffkQoUKyaucu67e9eVHdvl4ldXzPToavxxIS0tT+NYohW+NMq7jYG+vZg19s3QnuIdx6eo1szkF693z5UhuK+ZaWCWLm47muvf9vrxnaRUqVOiedTJ/v8+1c+Ce8/qRcqZ3fjb24Vkm0xDK0jHNqrtHbQIFASEUAJt4pJyHnuserOe6B0u6M+Hm4jWbNHbKbJM5kA7FHNe16zdV1LVwru3bs3SpXOsrO+7940m6M59RtlgeaFJg3L6ddN/lnmWy/7u59xtWOzvDPcuzPsdFViUnp+jFUZ9lK4CSZDZC6qU+XVSnuo9+nL9Sm3bsNRslkpaWpp17D2nn3kM6fe6iXur7hHHZj5+P0vzl6zX/13Dt3HvQ7I/X24lJWhm+TWs379Kiaf9V47o1svksbefdz77PVgB1R/6ZrDe7cvLa8DDnjrXkymucFVi6DMladT3oNe7fKuN9PT9Nqm0tmY0IzQ0fvz5QLRrV1k+Lf9OWyGiz8D05JcV4J7jvx72e61/i2ZKl93I7g/Xf77PC1uf1vZfYA/kdIRSAfMG9dEkN6t1F5y5c0ZSZpncpunnrdq6GUIUKZf7hw1IIEnvitFn7vcO3pTsTtOYmdwt3FIw/dc4sTIg9cSZX95sVHvcEedV8KuinSaMfuN2DPvjlhw+sWbFj70GdOHPepK1jW3/1f+pxlfNwMw7lD37mdZPLACxp0aiOWjSqI+nOpQGxJ87oaOxJLf3tD/2x4+/L3ib+sEAv9u4su/8/RnZ2durZ5VH17PKoUlJSdfzUWcWeOKODR49r1uI1+uv4KUl3Rkl9PXOxQr94S5LkUaaUyd3dMjt/7j3H7ezsVKZUCWMfd7t1O1FHjp3IldFQScnJ+nXdFpO2WlW9NfLFHqpcsbxxTp63x083mRPK2vLTa0OGnJ47eDBrvcYVZPee6waDQb//9IWKZeGDf9kydyaRt3Sn3OMnz5r1fdzC/6PcZun/dJyF18O0tDSdOG36eu9+1//pMm4lTOYtszSHWWpqqlkfOfV4G3893sZfaWlpij99XrEnzujIsXjNX77eGN6np6drwoxfrBZClSpeVE6ODiZffmT/iwPby7VzoFQJk78Jjls4B9LS0sz+brhbqeJF5ehgb3LX0+EDn1avJx69/5OQVMzVNuEbkFP/3HdKAPnKhctX9cKozxQZfeS+6928bfqtXqFCdip5z7fY916rf+/d4x6GpQlz/7dwjXnbotVmbf4NsjbZblbVq1lZ9vcMK5+/fL3Zenffnelh3HtcpTvhgiUtGtU2eXwk9qTOnr+sR8p5WPzxKuuuY8dPKzkl7y/FsYbT58wvS/ly9Ctq1aSuKlcsr0fKeejSlWv3DaBSU1PNLrks51FazRvW1rPdOuiHz940WXb56jVd+P8h/1evXTf53djbF1KlR8qpbfMG+s+zIRo95FmTbY/cFTo1v+c8jTt5VhvuubPTjZu3tGil6USrdWtWNgZALf8/+Ljb+G/nZPptcHYmJr905ZrJH+GS9PqLPdWxbTPVqPyIHinnoRJFXU3m9MpMxsS8GR7mtSI/vTY8zLlTEFm6RNfa3/wX1Ne47B6r7Kzf7J7zOD09XWs37870mDxSzkM3byXq5JkLxveXBrWrmfU7Z9nvZv3OC8vZXTWzo5xHabPLZ+eGrVPyPa8/S3/bbDbSsFn9WsZ/3/ucNu/8UyfvCRuWr93y0POh3byVqIS75pSys7NTxfIeCmhaTwN7dtKXowebrH/3676Uvff4B7Gzs5P/XcdAklas26p990wSniEnN6fIC7l3DlQ1WbZ55z6zLyRWhm+77x0SCxUqpKb3HNM1m3aoTKkSmf7/KlmsqLZHHTSbcgDI7xgJBSBPpKWmGe98V9W7gjoENFbD2tX1SHl3FXZx1sXLCVq1YZtCf1llsl2TejXN/kh2K1Hc5PGFS1f1v4Wr1aJRbeMfWTm940uFsmXUrmUjk7tgLVy5QUUKO///bdgLacmaP8wmHm7e0FfVK2U+8XBOFCnsosfbNNWy3/++tfH6LZEaOmaS+j7ZXoVdnDRr0ZpcuwV76ZLFzdqmzFykkPat5PT/dxkr515a9vaF9HibZnJ3m2m8VXd6err6DP1Ig3p3UasmdVW6VHHduHlbfx0/pZ37Dmll+DbFnzqnRd99ZJM5rHKbpWP14Vcz1a97sOztC2l71AF9Nm3effu4nZik+o89r9ZN66l1k7qqVdVbHmVKycnRQecvXtHMhavMtin8/5Mmb4s8oMGjJ6hdy0Zq3rC2qvqUl1uJ4rKzMyj2xBlN+nGhyXZFCjsb/92zy6P66seFJh80X3hzvEYPeU4N61TT6XMXNW7qHJPJpCWp//9fOivduQtavZqVTb75XvZ7hK4OHqOBPTupind53U5M0v4jsZoXtk5BrRppUO8u9z0eGYoXLSL7QoVM5ueZOnupihcrIne3kjp8LF4Tvv/FYhB4L7d7fk9/bN+r8C2R8vYqKzs7g5wdHU2+zb6f/PTa8DDnTkHkVrKYWdu0OWEa2LOTMRh1dyspZyfzeZ5yqqC+xpUo5io7Ozul3XVHxJkLV8uzTCnjiGK3EsVUpLBLttcv71lG7Vs31pqNO4zrfjp1juJOntUT7VuqvEdpJaek6OSZ89qzP0ZrNu1Q1P6jGvliD2OAVb2Sl+pUr6R9h/4OK+YuXavirkX0ZHBr3bx1W9/+tCxP7ownSf26P2Zyl8TjJ8+q+8vvadiAp+RepqS2Re7XR5P+Z7JN8aJF1PWx1sbHT3cM1NI1fxgfJ6ek6Kn/vK/RQ55TlYrlFRl9RO/dNX9bTsWdPKPHnn1dbZo3UMvGdVS90iNyL11C9oUK6dTZi/r6f4tN1i9yz//57LzHZ8XzPTuafIGRnJKiroPe1av9uinQ309FXQvr9LmLCt8apSWrN2n7su+y+5TzRG6cA90eDzT5Wy01NU1P/+d9vT+0n7y9PLVr32F9MDFUD9L/qce0afte4+M/Dx3TEwPf1ou9u6hGZS+5ODvp4uUEHTgap43b92jt5t0qXbKYuj8ecJ9egfyHEApAnjsSe0JHYk88cD2DwaBhA7qbtTeqW92s7fVPppo8PrtrSY7r+/j1gYr887DJKJb/LVxt9uEyQzHXwvp01Es53t/9jPpPH/22aafJCI55Yes0L2yd8fHdd8l6GHVrVpaDvb2SU/7+BvDzaT/r82k/Gx/vCPtOj5TzUGEXJ33x7n/Ub+RY493Krl67ofHfzdX47+Y+dC35XRO/mnIrWdxkTqiflvxmMiqtSGFnuRZx0fUbtzLtJyU1Vesidmfp7obNG/qa3Oku4fpNLVq1UYtWbXzgtnffDbKcR2mNHvKs3vnse2Pb1Ws3NPLjbzLdvk2z+iZ3JTIYDJrw3ivqMvBtk+e3Ydses1FV9+7/QVycndSmeX2TS+227zmgroNML4XyKF3ygZOxNqpb3eT4XLp6TT1e+cD4uHlDXy2e9nGWa8tPrw0Pc+4UNN4VPM3+v81atEazFv09Em3Rdx8ZL03MDQX1Nc7ZyVF1qvuYBMT33rTgq/dfVc8uj+Zo/Y9fH6jI6CPGybjT0tLMXvseZPTQ59Rj8BiTkZPf/rRM3/60zPj43iDaWp7v2VErwrdqx56DxratkfvV89UPMt3mkzdeNJmL8NEWDdW6aT1tvOu1LybulPqNGGuyXW68V99KTNKK9Vu1Yv3WB6577+tudt7js6JD6yZ6umMbzf/17xHaCddv6uMps/TxlFkm6xbLxSkVcltunAMdWjdWswa+2nLX3WdjT5xR/9c/NdnuQed1x7bNFNKhlZas3mRs23MgRoNHT8jWcwLyOy7HA5An7O0LmYzGeBBnJ0eNf/slBTarb7asZpWK6hDQJDfLM+FdwVMLv/tIVb0fPL+NVzl3/TL1w1wf6ZChcsXy+u7TkWZ3dspQsnhRTRwzxKzdMZP176dk8aLGieKzon3rxgr9/C25lTAfpWCJaxEXm00amttcnJ004b3BZpd7GZc7Oeq7T0aqZC4NkfcqW0ZfvDv4wSta0LJxHb3yXFeTtoE9O2nsmy9mel7dLaRDK80Y/4bZHYd8q/lo0XcfqXLFcplsmXMfv/7CfedRGj7waQX6+z2wn6ceD1QFzzK5Vld+em3Iqoc5d/ILOzs7DenfLc/3W1Bf44Za+PImt9Z/pJyHlkz7WLWrZ36X2bsVKmRnNr9WQNN6+nD4gEwn8C7jVkIfjXw+yzU9DCdHB82e+K7at278wHVdnJ00+YOhZqNODAaDpn48XLWqeme67YgXeqjuPXePy8n7dFb5VvPWe8P6mbRl9z0+Kya894r+80xIgZ5vLrfOgWljR9739X5Aj8fVxO/BNwiZ/MEQPd+jY5YnuC/nUTpL6wH5CSOhAOSJUiWK6eC6WdqyK1pbI/dr74EYxZ44o7MXLuvm7duyL1RIxYoWUZWK5dWycR317PyoKpTN/MPj9E9f15TQRQpbG6HYE2dyPK9BZmpWqaj18yZq6W9/6Nd1W7Vn/1FdvJKg9PR0lSxeVHVqVFJwQFN1fzwgVy8BsaRD6yZaO2eCJsz4RZu279WVhGvyKF1SbZs31PAXntbBo3Fm25SxMOw+Kz4aMUA+Xp76ZcUGHTkW/8Db3rdv3Vjbl32n+b+u19rNuxV9+JguX72m1NQ0FStaRN4VPFWvZmW1alxXbZrXN7udckHWoXUTrQgdp0k/LlTE7mglXLuh0qWKq0Wj2hrSv7uqV/LSW+Myv/ygsIuzVv3vM23fc0A7og4q5vgpXbx8VZeuXJOdnUGlShRT9UpeateykXqHtDO5tKJ103r6ZeoH2h51QDv3HtLJM+d18XKCrly7LkcHB3mWKaU61X3UJailOj3azOIfswOeflyPBTbVrEVrtHH7HsXEnlTC9ZtydnZUOXc3Na5XU726PKrG9TL/o7lezSra8PMkha2N0Mr1WxV1IEYXLl1VSkqKSpcqrkfKeahFozrq0Dp7oXHF8h76fc4ETZgxX79t3KEz5y+rmGth1alRSS/06qR2LRtpyPtfPbCfYkWLaPmPn+qL6T9rw7Y9OnPuotl8U9mVH14bHubcKahe6tNFpUsW16xFq7X/SKyu3biVJ3ekKoivcR3bNtOcSaM1bU6Y9h6I0ZVrN0wut3vY9at4l9eaWZ9r9YYdClsbocjoIzp38bJuJybJ9f8v26tVtaKaN6yt9q0bG29ocLcXe3eWX60qmjxzkXbuOagbt26rrLubglo10rDnn9KhmOO5cSiypEQxV82a8I7+2LFPv6wI1849B3XmwiXdvp2k4kWLqKpPBQU0radnunWw+FykO5e6rQgdp29/WqbFqzYq9sQZFXZxll+tKhrUu4vaNK+v2YvXmG2THVW9K2jZ959o+56D2rH3oOJOnNGFy1d15ep1OdgXUulSxVWrmo8eC2yq7o8FWJwDKrvv8Q9ib19I7w/rp2e7ddDsxb9pa2S0/oo7pWs3bqmwi5Pc3UqqXq3Katu8wUPtx9py4xxwL11Sq/73ub6ZtVhL1/yhuJNn5eLspNrVffT80x31eFt/PfniOw+sxdHBQZ+88YL6PRWsOUt+19bI/Yo9cUbXrt+Ug4O9Spcqrmo+FdSobg21bd5AfrWq5PLRAKzPkG7re0oCAB7Kax9O0Zylf0/sWrpUcf25JtSqt4kGAABZszUyWk8MNA0gfvrqXbVrmfVLlQHgn4KRUACQz701fprcShTTY4FNVaPyI8bLos6ev6TQBatMAihJ6tn5UQIoAADyyNrNu7Ro1UZ1DW6tJvVqGid1T0xK1rqI3Xp7/DST9b3KllGrJvVsUSoA2BwhFADkc6fOXNAPP6/QZ9/Nk6ODvYoXc1VycorF2z1Xr/yIxcncAQCAdSQmJWvBig1asGKDDAaDShRzlb19IV2+cs1sImr7QoX02dv/ydKcfADwT0QIBQAFSFJyivGuRPdq06y+Jo0ZYvwGFgAA5K309HRdvnrN4jKP0iU14b1X1Ka5+U1XAODfgjmhACCf23sgRqs3btfWyP2KP3VOF68kKDExWa5FXPRIOXfV962qJ4Nbyb++r61LBQDgX+fSlQSF/R6hzbv+1MGjcbpwOUFXE67L2dlRbiWKqXZ1H7Vr2UhPtG/5j7hRAAA8DEIoAAAAAAAAWJ2drQsAAAAAAADAPx8hFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqyOEAgAAAAAAgNURQsFMly5d1KVLF1uXAQAAAAAA/kHsbV0A8p+YmBhblwAAAPKRxKQkDXl3nCRp0n/flJOjo40rAgAABREjoQAAAAAAAGB1hFAAAAAAAACwOkIoAAAAAAAAWB0hFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqyOEAgAAAAAAgNURQgEAAAAAAMDqCKEAAAAAAABgdYRQAAAAAAAAsDpCKAAAAAAAAFgdIRQAAAAAAACsjhAKAAAAAAAAVkcIBQAAAAAAAKsjhAIAAAAAAIDVEUIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6QigAAAAAAABYHSEUAAAAAAAArI4QCgAAAAAAAFZHCAUAAAAAAACrI4QCAAAAAACA1RFCAQAAAAAAwOoIoQAAAAAAAGB1hFAAAAAAAACwOkIoAAAAAAAAWJ29rQsAAABAweHToqdS09JsXQYAAAXa2V1LbF2CTTASCgAAAAAAAFZHCAUAAAAAAACrI4QCAAAAAACA1RFCAQAAAAAAwOoIoQAAAAAAAGB1hFAAAAAAAACwOkKoAiw2NlYGg0GBgYG2LgUAAAAAAOC+CKEAAAAAAABgdYRQAAAAAAAAsDpCqLssWrRI/v7+Kly4sEqXLq2nnnpKR48e1ZgxY2QwGBQaGmpc19vbWwaDQenp6Zo8ebLq1aunwoULy8/Pz7hOSkqKpk6dqmbNmqlYsWJycXGRn5+fJk6cqJSUFLP9R0VF6Y033lDDhg1VpkwZOTk5qVKlSvrPf/6jU6dOmaw7ZswY+fj4SJI2bNggg8Fg/OnXr581Dg8AAAAAAECO2du6gPziq6++0rBhw2RnZ6fWrVvL09NT27ZtU5MmTdS5c+dMt3vppZf0448/KiAgQDVr1lRSUpIk6datW+rYsaPWr1+vUqVKyd/fX87Oztq2bZtee+01rV+/XosXL5ad3d854KeffqqFCxeqbt26atmypaQ7wdTUqVO1ZMkS7dy5U+XKlZMk+fn5qVu3blq4cKE8PDwUHBxs7CdjWwAAAAAAgPyCEErSX3/9pTfeeEOOjo5atWqV2rRpI+nOSKYXX3xRP/74Y6bbLlq0SJGRkfL19TVpHzlypNavX68ePXrou+++U/HixSVJ165dU8+ePbVs2TJNmzZNL730knGbQYMG6auvvpKHh4exLS0tTf/973/1/vvv691339UPP/wgSQoJCZGfn58WLlyoGjVqmIzSAgAAAAAAyG+4HE/SDz/8oKSkJD3zzDPGAEqS7O3t9eWXX8rV1TXTbd98802zAOrcuXOaPn26vLy89OOPPxoDKEkqWrSoZsyYIUdHR02dOtVkuzZt2pgEUJJkZ2en9957T+XLl9eyZcse5mma8fX1tfgTExOTq/sBAAAAAABgJJSkzZs3S5Keeuops2UlSpRQ+/bttWjRIovbdunSxawtPDxcycnJCg4OlouLi9lyT09PVa1aVfv27dOtW7dM1rl48aKWLVumP//8U1euXFFqaqokKTk5WRcvXtSlS5dUqlSpHD1PAAAAAAAAWyGEknT69GlJkpeXl8XljzzySKbbWloWGxsrSZo+fbqmT59+331funRJ5cuXlyTNnTtXL774oq5fv57p+teuXcu1ECo6Otpi+70juwAAAAAAAB4WIdRDcnZ2NmtLS0uTdGfy8Hr16t13eycnJ0lSXFyc8a52EydOVMeOHVW+fHnjKKnmzZtry5YtSk9Pz8XqAQAAAAAA8gYhlKSyZcvq0KFDio+PV61atcyWx8fHZ6u/ChUqSLpzl7rJkydnaZsVK1YoKSlJI0eO1NChQ82W//XXX9mqAQAAAAAAID9hYnJJLVq0kCQtXLjQbNnVq1e1Zs2abPXXpk0bFSpUSMuXL1dycnKWtrl8+bKkvwOsu23cuFFnz541a3d0dJR05y5+AAAAAAAA+RkhlKT+/fvL0dFR//vf/7Rx40Zje2pqqkaMGKFr165lq7/y5ctrwIABio2NVa9evSwGSEePHjUJvapVqyZJmj17tm7cuGFsP3nypF566SWL+yldurQcHBwUExNjnMAcAAAAAAAgPyKEklS5cmWNHz9eiYmJatOmjdq2batevXqpWrVqWrhwofr27Svp75FHWfHVV18pKChICxcuVOXKldWyZUv17t1bTzzxhKpWraqqVatq1qxZxvW7dOkiX19f7dy5U1WqVFH37t3VqVMnVatWTSVLllTz5s3N9uHo6Kjg4GCdOXNG9erV07PPPquBAwfqxx9/fPiDAgAAAAAAkIsIof7f0KFDtWDBAjVq1Ehbt27V6tWr5efnp23bthknH3dzc8tyfy4uLlq5cqVmzpyppk2b6sCBA1qwYIF27typMmXK6IMPPtD48eON6zs6OmrTpk16+eWX5ezsrOXLl+vAgQN69dVX9dtvv8nBwcHifr7//ns988wzunjxoubMmaMZM2Zow4YND3cwAAAAAAAAcpkhndut3Vdqaqrq1q2rAwcO6NSpU/L09LR1SVbn6+srSYqOjrZxJQAAID9ITErSkHfHSZLC1kcq9f/vBAwAAHLm7K4lti7BJhgJ9f9iYmJ05coVk7bExES98cYb2r9/vx599NF/RQAFAAAAAABgDfa2LiC/+OWXX/T++++rYcOG8vLyUkJCgvbs2aPTp0+rdOnSmjJliq1LBAAAAAAAKLAIof7fo48+qj179mjr1q3au3evUlJSVL58eb388st666235OXlZesSAQAAAAAACixCqP/XuHFjzZ0719ZlAAAAAAAA/CMxJxQAAAAAAACsjpFQAAAAyLJjm+fJydHR1mUAAIACiJFQAAAAAAAAsDpCKAAAAAAAAFgdIRQAAAAAAACsjhAKAAAAAAAAVkcIBQAAAAAAAKsjhAIAAAAAAIDV2du6AAAAABQcPi16KjUtzdZlAABywdldS2xdAv5lGAkFAAAAAAAAqyOEAgAAAAAAgNURQgEAAAAAAMDqCKEAAAAAAABgdYRQAAAAAAAAsDpCqCyIjY2VwWBQYGCgSXt4eLgMBoP69etnk7oAAAAAAAAKCkKoAmzMmDEyGAwKDQ21dSkAAAAAAAD3RQgFAAAAAAAAqyOEAgAAAAAAgNXlaQgVHx+vV155RZUrV5azs7NKlSqlTp06KSIiwrhOVFSUnJycVKpUKZ04ccKsjxdeeEEGg0GDBg0yaU9PT9fcuXMVFBQkNzc3OTs7y9vbW08//bTWrl2bo1oe1qpVq9SxY0eVKVNGTk5OqlSpkoYPH66LFy+arduvXz8ZDAaFh4dr48aNatu2rYoWLapixYqpY8eO2r9/v8n63t7e+uCDDyRJ/fv3l8FgMP6Eh4fn2nMAAAAAAADIDXkWQm3ZskX16tXT119/LQcHB3Xs2FG1a9fW6tWr1bp1a/3888+SJD8/P3388ce6fPmynnvuOaWnpxv7WLJkib7//ntVq1ZNX375pbE9NTVVPXr0UO/evbVx40bVq1dPTz75pCpUqKBff/1VkydPzlEtD2PUqFF67LHH9Pvvv6t69erq0qWL7O3tNWHCBDVt2lRnz561uF1YWJjatm2rmzdv6vHHH1fZsmW1YsUKtW7dWmfOnDGu1717d9WrV0+S1KJFCz333HPGH09Pz4euHwAAAAAAIDfZ58VOEhIS1K1bNyUkJGj27Nnq06ePcdnOnTvVvn17DRw4UG3btlWZMmU0YsQIrVy5UuvWrdMXX3yhkSNH6vTp03rhhRfk4OCgn376SUWKFDH2MXbsWP3yyy+qVauWli9fLh8fH+Oyq1evKioqKse15MQvv/yicePGqXbt2lq8eLGqVKki6c5orTFjxujDDz/U0KFDNW/ePLNtJ06cqIULFyokJETS3wHbwoUL9c033+jDDz+UJH3++ecaM2aM9uzZo4EDB3KHPgAAAAAAkK/lyUioH374QadPn9awYcNMQh9JatSokUaPHq3r169r9uzZkiSDwaCZM2eqZMmSeueddxQZGal+/frpwoULev/999WoUSPj9klJSfriiy+M+7k7gJKk4sWLKyAgIMe15MTHH38sSZo7d64xgMp4XmPGjJGfn58WLFigCxcumG3bq1cvYwAlSYUKFdJbb70lSdq4cWOOa7LE19fX4k9MTEyu7gcAAAAAACBPQqg1a9ZIkrp27WpxeatWrSRJ27dvN7ZVqFBB3333nZKSkhQYGKg1a9aoZcuWGjVqlMm2O3fu1JUrV1SvXj01bdrUKrVkx7lz57Rnzx5VrVpVtWvXNltuMBjUokULpaamateuXWbL27dvb9ZWrVo1SdLp06dzVBMAAAAAAICt5cnleLGxsZLuzF10P/eODHrqqaf05JNPavHixSpcuLBmzZqlQoUKmawTHx8vSapcubJVa8mqjP6PHDkig8GQ7X1UqFDBrK1o0aKSpMTExBzVlJno6GiL7b6+vrm6HwAAAAAAgDwJodLS0iTdmUz77rmc7lWjRg2Tx6dPn9amTZskSTdv3tT+/fvl7e1tk1qy27+np6c6dOhw33UrVqxo1mZnl6c3LAQAAAAAAMgTeRJCVahQQYcOHdKoUaPUsGHDLG2Tnp6u/v3768KFC+rVq5fmz5+vAQMGaN++fSYThnt5eUlSlucxykkt2ZExkql06dIKDQ3N9f4BAAAAAAAKojwZdhMUFCRJWrx4cZa3mTx5slavXq0WLVpo1qxZeuedd3T27Fk9//zzJus1bNhQJUqU0J49e7I0j1NOasmOChUqqEaNGtq/f78OHz5slX1kcHR0lCSlpKRYdT8AAAAAAAAPK09CqEGDBsnd3V3jx4/XtGnTjJesZUhJSdHq1av1559/SpL279+vN998U0WLFjXOAzV69Gg1adJEYWFh+vbbb43bOjk56bXXXpMkPf/884qLizPp++rVq9qwYUOOa8mJ0aNHKy0tTd26dVNUVJTZ8osXL2r69Ok57j9DuXLlJEmHDh166L4AAAAAAACsKU8uxytRooSWLl2qzp07a9CgQfrvf/+r2rVrq2TJkjpz5ox2796tK1euaPHixapWrZr69Omj27dv67vvvpOPj8+dQu3tNXv2bNWvX18jRoxQmzZtVL16dUnS22+/rcjISC1ZskTVqlVTq1at5O7urvj4eO3evVtBQUEKCAjIdi2W7m6XFb1791Z0dLQ++eQTNWzYUH5+fqpcubLS09MVExOjvXv3ytXVVS+88MJDHdf27dvL2dlZEyZM0J9//qly5crJYDDo9ddfNx4bAAAAAACA/CDPZsH29/fXvn379MYbb6hYsWLasGGDlixZori4OAUEBCg0NFTt2rXTu+++q6ioKD311FN69tlnTfqoWrWqvvzyS928eVN9+vRRcnKypDsB1cKFCxUaGip/f3/t3LlTixYt0okTJ9SpUycNGzYsR7U8jI8//lgbNmxQt27ddObMGS1ZskTr169XamqqXn75ZS1btuyh+pfujIRaunSp/P399ccff+iHH37QjBkzdPr06YfuGwAAAAAAIDcZ0tPT021dBPIXX19fSVJ0dLSNKwEAAPlBYlKShrw7TpIUtj5SqfdMZwAAKJjO7lpi6xLwL5NnI6EAAAAAAADw70UIBQAAAAAAAKsjhAIAAAAAAIDVEUIBAAAAAADA6gihAAAAAAAAYHX2ti4AAAAABcexzfPk5Oho6zIAAEABxEgoAAAAAAAAWB0hFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqyOEAgAAAAAAgNURQgEAAAAAAMDq7G1dAAAAAAoOnxY9lZqWZusyAKDAO7tria1LAPIcI6EAAAAAAABgdYRQAAAAAAAAsDpCKAAAAAAAAFgdIRQAAAAAAACsjhAKAAAAAAAAVkcIlQtu3LihIUOGyMvLS/b29jIYDBozZoytywIAAAAAAMg3CKFywVtvvaXJkyfL2dlZTz/9tJ577jn5+fnlyb779esng8Gg8PDwPNkfAAAAAABATtjbuoB/giVLlsjFxUWRkZFydXW1dTkAAAAAAAD5DiOhcsGJEyfk7u5OAAUAAAAAAJCJAhtCxcbGymAwKDAwUAkJCRo6dKi8vLzk7OysmjVrasKECUpLSzPb7ubNmxo7dqzq168vV1dXubq6yt/fXzNnzrS4H4PBIG9vbyUlJenDDz9UjRo15OTkpJCQEAUGBspgMCg9PV1xcXEyGAzGn7tdunRJb731lmrVqiUXFxcVL15cbdu21fLlyzN9fvHx8RoyZIiqVasmFxcXlSpVSo0aNdIHH3yghIQEY20Zdbdp08Zk/7GxsTk8sgAAAAAAALmvwF+Ol5iYqLZt2yomJkZt27ZVUlKS1q5dq+HDh2vPnj0KDQ01rnvu3DkFBQVp79698vT0VEBAgNLT0xUREaF+/fpp586dmjx5stk+0tLSFBISoo0bNyogIEB169aVm5ub/P395e3trZkzZ6pIkSLq3r272baHDx9Wu3btFB8fL29vb3Xo0EHXrl3T1q1b1blzZ3322WcaOXKkyTabNm1Sly5ddOXKFXl7e6tz5866deuWDh48qDFjxuiJJ56Qn5+fnnvuOf3xxx+KiYlRhw4d5OnpaeyDUVkAAAAAACA/MaSnp6fbuoiciI2NlY+PjySpbt26Wrt2rUqXLi1JiomJUevWrXXq1CktXrxYISEhkqSOHTtqxYoVGjp0qMaNGycnJydJ0tmzZ9WpUyft3LlTK1euVHBwsHE/GaOaqlSpovDwcJUvX96sFoPBoIoVK5qNPkpNTVX9+vW1b98+jR8/XiNGjJCd3Z3BZ0ePHlX79u11/PhxRUVFqXbt2pLujJqqUaOGzp8/r88++0zDhw83biNJW7ZsUeXKleXu7i7pzsTkM2fO1Pr16xUYGJitY+jr62uxPSYmRpUrV1Z0dHS2+gMAAP9MiUlJGvLuOElS2PpIpVoYbQ4AyJ6zu5bYugQgzxXYy/Hu9vnnnxsDKEmqXLmyRo8eLUmaMmWKJCkqKkorVqxQ48aN9eWXXxoDKEny8PDQtGnTJElTp061uI+xY8daDKDuJywsTPv27VO3bt30+uuvm4RJVapU0RdffKHU1FRNnz7d2P7999/r/PnzCg4O1siRI022kaRmzZoZAygAAAAAAICCosBfjleqVCkFBQWZtffq1Usvv/yyIiIilJaWpjVr1kiSQkJCzIIdScY5orZv3262zGAwqHPnztmuLWOfXbt2tbi8VatWkmSyz99//12SNGjQoGzvL7syG+mU2QgpAAAAAACAnCrwI6EqVqxosb148eIqUaKEbt26pcuXLxsvlXvnnXdMJvC+++f69eu6cOGCWV/u7u4mI6eyKmOfffr0sbi/MmXKSJLJPuPj4yXdGc0FAAAAAADwT1HgR0JlVcad8lq2bJntgMfZ2fmh9hkcHCwPD49M17v7UkIAAAAAAIB/ogIfQh0/ftxie0JCgq5cuSIXFxeVKFFCFSpUkHTncrwRI0bkSW0Z+xw4cKC6deuWpW28vLx08OBBxcTEqE6dOtYsDwAAAAAAIM8U+MvxLl68qLVr15q1z5s3T9KdibwLFSpknDdq8eLFeVZbTvbZrl07STJOlP4gjo6OkqSUlJRsVgcAAAAAAJB3CnwIJUkjR47UxYsXjY+PHTumDz/8UJI0ePBgSVLTpk0VFBSkzZs3a/DgwUpISDDrZ8+ePVq1alWu1dWtWzfVqlVLP/30kz766CMlJiaaLE9PT9fmzZu1efNmY9vAgQNVunRprVy5UhMnTlR6errJNlu3btW5c+eMj8uVKydJOnToUK7VDQAAAAAAkNsM6femHAVEbGysfHx85O/vr6SkJP31119q27atkpOTtXbtWt28eVN9+/bVrFmzjNucO3dOwcHBioyMVIkSJeTn56dy5crp6tWr2rt3r+Lj4zV06FBNnDjRuI3BYFDFihWNk4xbcr91jhw5og4dOujYsWNyd3dX3bp15e7urgsXLigqKkrnzp3ThAkTNGzYMOM24eHh6tKli65duyYfHx81btxYt27d0oEDB3T06FFFRkbKz89PkrRr1y41btxYTk5O6tChg3F+qXHjxsnNzS1Hxzbj7niZ3T0PAAD8uyQmJWnIu+MkSWHrI5X6//NeAgBy7uyuJbYuAchzBX5OKCcnJ61atUpvv/22lixZogsXLsjHx0cvvPCCSbAj3bnLXUREhKZPn6558+YpMjJSERER8vDwUKVKlTRkyBD17NkzV+urWrWqIiMjNWXKFC1atEhbt25VSkqKPD09Vb9+fXXp0kVPP/20yTaBgYHas2ePxo8fr1WrVmnJkiVydXWVj4+PPvzwQ5OJ1Rs2bKjZs2friy++0Jo1a3Tr1i1J0rvvvpvjEAoAAAAAACC3FfiRUAEBAQoPD7d1Of8ojIQCAAB3YyQUAOQ+RkLh3+gfMScUAAAAAAAA8jdCKAAAAAAAAFgdIRQAAAAAAACsrsBOTO7t7a0COp0VAAAAAADAv06BDaEAAACQ945tnicnR0dblwEAAAogLscDAAAAAACA1RFCAQAAAAAAwOoIoQAAAAAAAGB1hFAAAAAAAACwOkIoAAAAAAAAWB0hFAAAAAAAAKzO3tYFAAAAoODwadFTqWlpti4DAHR21xJblwAgmxgJBQAAAAAAAKsjhAIAAAAAAIDVEUIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6QigAAAAAAABYHSHUXcaMGSODwaDQ0NA82d+NGzc0ZMgQeXl5yd7eXgaDQWPGjMlxf4GBgTIYDIqNjc21GgEAAAAAAHKDva0L+Dd76623NHnyZFWpUkVPP/20HB0d5efnZ+uyAAAAAAAAch0hlA0tWbJELi4uioyMlKurq63LAQAAAAAAsBoux7OhEydOyN3dnQAKAAAAAAD84z10CHX79m05OzvL29vbbFlISIgMBoNatmxptqxRo0ays7PT+fPnjW379+9Xnz59VLZsWTk6Oqp8+fJ69tlndejQIbPtw8PDZTAY1K9fP126dEkvv/yyypYtKycnJ9WuXVs//PBDpjUvW7ZMzZo1U+HCheXm5qZu3brp8OHD932eN2/e1NixY1W/fn25urrK1dVV/v7+mjlzpsX1DQaDvL29lZSUpA8//FA1atSQk5OTQkJCjHM3paenKy4uTgaDwfgjSbGxsTIYDAoMDLTYd17PXQUAAAAAAPCwHvpyPGdnZzVt2lQbN25UbGysMYxKS0vTxo0bJUk7duzQzZs3VbhwYUnS1atXFRkZqVq1aqlMmTKSpLVr16pz5866deuW6tevr8DAQB08eFCzZs3S4sWLtWLFCrVq1cps/1euXFGzZs10/fp1tWrVShcuXNDGjRv1/PPPKy0tTQMHDjRZ/9tvv9XLL78sg8GgVq1aqWzZstq6dauaNGmizp07W3yO586dU1BQkPbu3StPT08FBAQoPT1dERER6tevn3bu3KnJkyebbZeWlqaQkBBt3LhRAQEBqlu3rtzc3OTv7y9vb2/NnDlTRYoUUffu3XN8/AEAAAAAAAqCXJkTKjAwUBs3blR4eLj69esnSdqzZ48uX74sX19fRUdHKyIiQu3atZMkbdy4UWlpacaRPjdu3FCfPn1069YtTZkyRYMHDzb2PWHCBA0fPly9e/fWkSNH5OzsbLLvpUuXqmfPngoNDZWTk5OkO3MtPfnkk/roo49MQqi4uDi99tprcnBwUFhYmDp06CBJSk5OVv/+/TV79myLz69///7au3evhg4dqnHjxhn3c/bsWXXq1ElTpkxRx44dFRwcbLJdfHy8nJycdOjQIZUvX96s35kzZ6p06dI2G9Hk6+trsT0mJkaVK1fO42oAAAAAAMA/Wa7MCZURJoWHhxvbMv793nvvZbosICBAkjR//nydPXtWzZo1MwmgJOm1115Tw4YNdeLECS1cuNBs38WKFdOUKVOMwZB05zLA2rVr6/jx44qNjTW2//DDD7p9+7Z69eplDKAkycHBQV999ZVxpNbdoqKitGLFCjVu3FhffvmlyX48PDw0bdo0SdLUqVMtHpuxY8daDKAAAAAAAAD+TXIlhPL395eTk5NZ0FS0aFF169ZNFStWtBhCZYRXmzZtkiT16dPHYv99+/Y1We9uDRs2lJubm1l7tWrVJEmnT582tmVs37NnT7P13dzc1L59e7P2NWvWSLoTbNnZmR+ujDmitm/fbrbMYDBkeolffhAdHW3xh1FQAAAAAAAgt+VKCOXi4qImTZooLi5OsbGxSktL06ZNm9SqVSsVKlRIgYGBxnmhrl69qqioKJP5oE6dOiVJFic3v7v95MmTZssqVKhgcZuiRYtKkhITE41tGfupWLHiffdzt4yRVO+8847JBOJ3/1y/fl0XLlww29bd3d1k5BQAAAAAAMC/Va7MCSXdGdW0adMmhYeHq169erp8+bJxpFNgYKBmzpypiIgI3bp1S2lpacZL8bIi465xllganZSb0tLSJEktW7bM9gihe+evyi0ZNQEAAAAAABQUuRZCBQQE6KOPPlJ4eLguX74sSSYhlHTnMrxbt26ZtElSuXLlJN2ZONySjNFIDzu3UtmyZXXo0CHFxcWpVq1aZsst7T9jpFVISIhGjBjxUPvPKkdHR0nS9evXLS6Pj4/PkzoAAAAAAAByS64NI2revLkcHR0VHh6u8PBwFStWTA0aNJB05zK3jHmh7p2UXJJatWolSZo7d67FvjPuWpexXk5lbD9//nyzZZcuXTLO/3S3oKAgSdLixYsfat/ZUbp0adnb2+vYsWNKSUkxWZacnKwNGzbkWS0AAAAAAAC5IddCqLvnhVqzZo1xPqgMgYGB2r59u6KiolSjRg15eHgYlz399NPy8PDQH3/8YbzbXIZJkyZp586dKl++vLp16/ZQNfbv319OTk766aef9Pvvvxvbk5OT9dprr+nGjRtm2zRt2lRBQUHavHmzBg8erISEBLN19uzZo1WrVj1UbXdzdHRUs2bNdOnSJX399dfG9pSUFI0YMULHjh3LtX0BAAAAAADkhVydUCljdNPt27dNLreT7oRQycnJSktLM1tWpEgR/fTTT3JxcdGgQYPUqFEj9e7dWw0aNNDQoUPl6uqquXPnPvQcSz4+Pvriiy+UnJysDh06qE2bNurVq5eqVaumpUuXZnp3vtmzZ6t+/fr65ptvVLFiRbVp00Z9+vRRp06d9Mgjj8jPzy9XQyhJeu+992RnZ6dhw4apefPm6tq1qypXrqy5c+fqueeey9V9AQAAAAAAWFuuhlB3h0uWQqjMlknSo48+qh07dqhXr146ceKEFixYoDNnzqhv377auXPnQ1+Kl2Hw4MFavHixGjdurG3btmn16tWqV6+etm7dqipVqljcxt3dXREREZo0aZJq1aqlyMhILViwQHv37lWlSpX02WefaeTIkblSX4Z27dpp2bJlaty4sXbv3q0NGzbI399fO3bsyPQuggAAAAAAAPmVIT09Pd3WRSB/8fX1lSRFR0fbuBIAAJAfJCYlaci74yRJYesjlcqdegHkA2d3LbF1CQCyKVdHQgEAAAAAAACWEEIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6e1sXAAAAgILj2OZ5cnJ0tHUZAACgAGIkFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqyOEAgAAAAAAgNURQgEAAAAAAMDqCKEAAAAAAABgdfa2LgAAAAAFh0+LnkpNS7N1GQD+Bc7uWmLrEgDkMkZCAQAAAAAAwOoIoQAAAAAAAGB1hFAAAAAAAACwOkIoAAAAAAAAWB0hFAAAAAAAAKyOEMqKQkNDZTAYNGbMGFuXAgAAAAAAYFOEUAAAAAAAALA6e1sX8E/25JNPyt/fX6VLl7Z1KQAAAAAAADZFCGVFxYsXV/HixW1dBgAAAAAAgM3Z/HK8X3/9VQMGDFDNmjVVrFgxFSlSRPXq1dMnn3yixMREk3XvnmMpJiZGTz/9tEqXLq1ixYrpscce0/79+yVJKSkp+uSTT1StWjU5OzurSpUq+vrrrx96/7GxsTIYDPf9CQwMtFjv3fr16yeDwaDw8HBt3LhRbdu2VdGiRVWsWDF17NjR+DzudePGDY0aNUre3t7G5/XRRx8pOTlZ3t7eMhgM2Tz6AAAAAAAAecPmI6Gef/553bp1S7Vr11bdunV19epVbd++Xe+8847Wrl2rNWvWqFChQibbHDt2TE2aNJGHh4fatWun/fv3a9WqVdq1a5f27t2rl156SeHh4WrTpo0qVaqk9evX65VXXpGjo6NeeOGFHO/f1dVVzz33nMXnERUVpT179pjVej9hYWH66quv1KhRIz3++OOKiorSihUrtG3bNv3555/y9PQ0rpuYmKh27dpp69atKlWqlDp16qTExER9+umn2r17d5b3CQAAAAAAYAs2D6G+++47tW/fXi4uLsa2a9euqXfv3lq+fLl++uknPfvssybb/O9//9OoUaP0ySefyGAwKD09XQMGDFBoaKgeffRR2dnZ6ciRIypTpowkae3atWrXrp0+/vhjsxAqO/svXbq0QkNDzZ5DTEyMmjRpIkdHR3344YdZfu4TJ07UwoULFRISIklKTU1Vjx49tHDhQn3zzTcmfX322WfaunWrmjRpotWrV6tEiRKS7ozOat26teLj47O8XwAAAAAAgLxm88vxnnjiCZMASJKKFi2qCRMmSJKWLl1qtk2lSpX04YcfGi8/MxgMeu211yRJ+/fv18SJE40BlCQ9+uijql+/vuLi4hQbG/vQ+79bQkKCunTpokuXLmnq1Klq0aJFFp71Hb169TIGUJJUqFAhvfXWW5KkjRs3mqz77bffSpK++OILYwAlSd7e3nrvvfeyvM+7+fr6WvyJiYnJUX8AAAAAAACZsflIKEk6cuSIVqxYoaNHj+rGjRtKS0tTenq6cdm9AgMD5eDgYNJWqVIlSZKDg4PJvEx3L4+MjNTp06fl7e39UPvPkJaWpt69e2v//v0aNmyYBgwYkJ2nrfbt25u1VatWTZJ0+vRpY1tcXJxOnjwpT09PtWzZ0mybHj16mI3wAgAAAAAAyE9sGkKlp6dr5MiRmjBhgjH0ude1a9fM2sqXL2/W5urqKkny9PS0OC9TxvK7JxvP6f4zjBo1Sr/++qvat2+vzz//PNP1MlOhQgWztqJFi5rVmRFIeXl5WeynaNGiKlGihK5cuZKt/UdHR1ts9/X1zVY/AAAAAAAAD2LTy/F+/vlnffnll6pQoYIWLFigkydPKikpSenp6cYQxlI4ZGeXedn3W5Zb+5ekWbNm6bPPPlO1atX0888/Z2tC8pzUCgAAAAAAUJDZdCTU4sWLJUlTp05Vx44dTZb99ddf+Xb/27Zt0wsvvKASJUpo2bJlJnM0WUPZsmUlKdPJx69du5btUVAAAAAAAAB5yaZDcS5fvizJ8mVp8+fPz5f7P3HihEJCQpSSkqKff/5Z1atXt2qNklSxYkWVL19eZ86cUUREhNnyX375xeo1AAAAAAAAPAybhlAZk3BPmzbN5LK3TZs26bPPPst3+79165ZCQkJ05swZff755xYnFreWl156SZI0YsQIXb161dgeFxenDz/8MM/qAAAAAAAAyAmbXo43ZMgQhYaG6ptvvlF4eLjq1q2rkydP6o8//tCIESNyNNm3Nfe/YMEC7dq1S66uroqKilK/fv3M+qxRo4ZGjRqV67W+/vrr+vXXX7V161ZVrlxZbdq0UWJiotatW6dHH31U6enpJnfUAwAAAAAAyE9sPhJq586d6ty5sy5cuKBly5bp+vXr+u677/JsJFR29p+amipJun79umbOnGnxZ9WqVVap1cnJSb/99pveeOMNFSlSRMuWLVN0dLRGjBihn3/+WWfPnpWbm5tV9g0AAAAAAPCwDOmZ3f4NBcbWrVvVrFkzBQcHa+XKlQ/dn6+vryQpOjr6ofsCAAAFX2JSkoa8O06SFLY+UqlpaTauCMC/wdldS2xdAoBcZtORUMieyMhIpd3zR99ff/2lQYMGSZL69u1ri7IAAAAAAAAeyKZzQiF7evXqpatXr6pOnTpyc3PT8ePHtWvXLiUmJqpLly7q3bu3rUsEAAAAAACwiBCqAHn11Vc1f/587d27V5cuXZKzs7P8/PzUt29fvfTSSzIYDLYuEQAAAAAAwCJCqAJk8ODBGjx4sK3LAAAAAAAAyDbmhAIAAAAAAIDVMRIKAAAAWXZs8zw5OTraugwAAFAAMRIKAAAAAAAAVkcIBQAAAAAAAKsjhAIAAAAAAIDVEUIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6e1sXAAAAgILDp0VPpaal2boMAHng7K4lti4BwD8MI6EAAAAAAABgdYRQAAAAAAAAsDpCKAAAAAAAAFgdIRQAAAAAAACsjhAKAAAAAAAAVkcIlU2xsbEyGAwKDAy0dSkAAAAAAAAFBiFUAUcoBgAAAAAACgJCKAAAAAAAAFgdIRQAAAAAAACsjhDqISQkJGjo0KHy8vKSs7OzatasqQkTJigtLc1s3Zs3b2rs2LGqX7++XF1d5erqKn9/f82cOdNi33FxcXr55ZdVrVo1FS5cWKVKlZKvr68GDRqkQ4cOSZLGjBkjHx8fSdKGDRtkMBiMP/369bPa8wYAAAAAAMgue1sXUFAlJiaqbdu2iomJUdu2bZWUlKS1a9dq+PDh2rNnj0JDQ43rnjt3TkFBQdq7d688PT0VEBCg9PR0RUREqF+/ftq5c6cmT55sXD8+Pl4NGjTQpUuXVLVqVT3++ONKTU1VXFycpk+frmbNmql69ery8/NTt27dtHDhQnl4eCg4ONjYR8uWLfPycAAAAAAAANwXIVQObd26VXXr1tWRI0dUunRpSVJMTIxat26tmTNnKiQkRCEhIZKk/v37a+/evRo6dKjGjRsnJycnSdLZs2fVqVMnTZkyRR07djSGSN9//70uXbqkV155xSSckqTjx48rOTlZkhQSEiI/Pz8tXLhQNWrUMAm+AAAAAAAA8hMux3sIn3/+uTGAkqTKlStr9OjRkqQpU6ZIkqKiorRixQo1btxYX375pTGAkiQPDw9NmzZNkjR16lRj+/nz5yVJ7dq1M9vnI488osqVK+dK/b6+vhZ/YmJicqV/AAAAAACADIRQOVSqVCkFBQWZtffq1UuSFBERobS0NK1Zs0bSnVFLdnbmhztjjqjt27cb2xo2bChJevvtt7V8+XLdvn3bGk8BAAAAAAAgz3A5Xg5VrFjRYnvx4sVVokQJXblyRZcvX1ZsbKwk6Z133tE777yTaX93B039+vXTmjVrNH/+fHXu3FnOzs5q3LixgoODNWDAAHl6eubKc4iOjrbY7uvrmyv9AwAAAAAAZCCEsrKMO+W1bNkyy5fRFSpUSD///LNGjRqlpUuXat26ddq2bZs2bdqkTz/9VKtWrVLz5s2tWTYAAAAAAECuIoTKoePHj1tsT0hI0JUrV+Ti4qISJUqoQoUKku5cjjdixIhs7aN+/fqqX7++xowZo4SEBI0ZM0YTJkzQsGHDTC7fAwAAAAAAyO+YEyqHLl68qLVr15q1z5s3T5LUrFkzFSpUyDhv1OLFix9qf8WKFdPYsWNlMBj0559/GtsdHR0lSSkpKQ/VPwAAAAAAgDURQj2EkSNH6uLFi8bHx44d04cffihJGjx4sCSpadOmCgoK0ubNmzV48GAlJCSY9bNnzx6tWrXK+HjWrFkmQVOGlStXKj09XV5eXsa20qVLy8HBQTExMUpNTc215wYAAAAAAJCbuBwvh/z9/ZWUlKQqVaqobdu2Sk5O1tq1a3Xz5k317dtXXbt2Na47e/ZsBQcH65tvvtGcOXPk5+encuXK6erVq9q7d6/i4+M1dOhQBQcHS5IWLlyoZ599VpUrV1adOnXk4uKiY8eOadu2bbKzs9N///tfY9+Ojo4KDg5WWFiY6tWrpwYNGsjR0VEtWrRQ//798/y4AAAAAAAAWEIIlUNOTk5atWqV3n77bS1ZskQXLlyQj4+PXnjhBQ0bNsxkXXd3d0VERGj69OmaN2+eIiMjFRERIQ8PD1WqVElDhgxRz549jesPHz5cFSpU0ObNm7Vp0ybduHFD5cqVU48ePTRixAg1atTIpP/vv/9eI0eO1G+//aY5c+YoNTVVKSkphFAAAAAAACDfMKSnp6fbugjkL76+vpKk6OhoG1cCAADyg8SkJA15d5wkKWx9pFL//+6/AP7Zzu5aYusSAPzDMCcUAAAAAAAArI4QCgAAAAAAAFZHCAUAAAAAAACrI4QCAAAAAACA1RFCAQAAAAAAwOrsbV0AAAAACo5jm+fJydHR1mUAAIACiJFQAAAAAAAAsDpCKAAAAAAAAFgdIRQAAAAAAACsjhAKAAAAAAAAVkcIBQAAAAAAAKsjhAIAAAAAAIDV2du6AAAAABQcPi16KjUtzdZlALDg7K4lti4BAO6LkVAAAAAAAACwOkIoAAAAAAAAWB0hFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqytQIZTBYJC3t7fN9h8aGiqDwaAxY8bYrAYAAAAAAICCKN+EULGxsTIYDAoMDLR1KQUKwRgAAAAAACgI8k0IBQAAAAAAgH8uQigAAAAAAABYXb4IocaMGSMfHx9J0oYNG2QwGIw//fr1M1s/NTVV48aNU7Vq1eTk5CQvLy+9+eabSkxMtNj/zZs3NXbsWNWvX1+urq5ydXWVv7+/Zs6cmeOaDx06pG7dusnNzU1FihRRixYttGLFikzXj4+P1yuvvKLKlSvL2dlZpUqVUqdOnRQREWFx/YiICIWEhKhixYpycnKSp6enmjRpolGjRun69euSpMDAQPXv31+S9MEHH5gct9DQ0Bw/NwAAAAAAgNxmb+sCJMnPz0/dunXTwoUL5eHhoeDgYOOyli1bmq3fu3dvrVixQoGBgapevbo2bdqk8ePH6+TJk5o9e7bJuufOnVNQUJD27t0rT09PBQQEKD09XREREerXr5927typyZMnZ6vemJgYNW3aVKVKlVL79u116tQpbdq0SZ06ddKMGTOMwVCGLVu2qGPHjrp8+bKqV6+ujh076vz581q9erVWrVqln376ST169DCuHxYWppCQEKWnp6tJkyZq3ry5rly5oiNHjmjcuHF66aWX5OrqquDgYKWkpGjz5s2qV6+e/Pz8jH1UqVIlW88JAAAAAADAmgzp6enpti5CujMxuY+PjwICAhQeHm5xHYPBIEmqWbOm1q1bJ09PT0nSsWPH1KBBA125ckVHjx5V5cqVjdt07NhRK1as0NChQzVu3Dg5OTlJks6ePatOnTpp586dWrlypUnwlZnQ0FBjwPTss89qxowZsre/k+MtX75cISEhcnJy0uHDh1W+fHlJUkJCgmrUqKFz585p5syZ6tOnj7G/nTt3qn379kpOTtZff/2lMmXKSJICAgK0ceNGLViwQN26dTOpYceOHapRo4aKFi1qUtP777+fa5OT+/r6SpKio6NzpT8AAFCwJSYlaci74yRJYesjlZqWZuOKAFhydtcSW5cAAPeVLy7Hy65JkyYZAyhJ8vHxUd++fSVJmzZtMrZHRUVpxYoVaty4sb788ktjACVJHh4emjZtmiRp6tSp2dq/q6urJk6caAygJKlTp07q3r27bt68qR9//NHY/sMPP+j06dMaNmyYSQAlSY0aNdLo0aN1/fp1kxFc58+flyS1a9fObN+NGzc2BlAPy9fX1+JPTExMrvQPAAAAAACQocCFUA4ODmrTpo1Ze7Vq1SRJp0+fNratWbNGkhQSEiI7O/OnmjFH1Pbt27NVQ/v27VWyZEmz9l69ekkyDcIyaujatavFvlq1aiVJJjU0bNhQkvTMM89ox44dSuPbRgAAAAAAUMDlizmhssPT01OFChUya88YHXT35OSxsbGSpHfeeUfvvPNOpn3evn07WzVUrFjRYru3t7ck6dSpU2Y1tGjR4r59XrhwwfjvTz75RPv27VNYWJjCwsJUsmRJtWzZUl26dFHfvn3l7OycrXozk9nldhmX4wEAAAAAAOSWAhdCWRrRlJmMEUQtW7Y0mScqL2XU0L17dxUpUiTT9WrUqGH8t5eXl3bu3Kl169Zp+fLl2rBhgzGQGj9+vLZs2SI3Nzer1w4AAAAAAJBbClwIlR0VKlSQdOdyvBEjRuRav3FxcfdtL1eunEkNhw4d0qhRo4yX2WWFvb292rdvr/bt2xv7HjBggNatW6dx48Zp/PjxD/EMAAAAAAAA8la+mRPK0dFRkpSSkpJrfQYFBUmSFi9enGt9Snfmebpy5YpZ+7x58yTdGXmV2zVUrFhRb775piTpzz//NLZb47gBAAAAAADktnwTQpUuXVoODg6KiYlRampqrvTZtGlTBQUFafPmzRo8eLASEhLM1tmzZ49WrVqVrX6vX7+u4cOHmwQ/K1eu1Pz58+Xi4qL+/fsb2wcNGiR3d3eNHz9e06ZNM5tkPCUlRatXrzYJliZMmKAzZ86Y7XfFihWS7lyulyFj1NWhQ4ey9RwAAAAAAADyUr65HM/R0VHBwcEKCwtTvXr11KBBAzk6OqpFixYmoU52zZ49W8HBwfrmm280Z84c+fn5qVy5crp69ar27t2r+Ph4DR06VMHBwVnus0+fPlq0aJHCw8PVtGlTnT59Whs3blR6eromTZpkvAxQkkqUKKGlS5eqc+fOGjRokP773/+qdu3aKlmypM6cOaPdu3frypUrWrx4sWrXri1J+uCDDzRy5EjVq1dPVatWVXp6uvbs2aPDhw+rVKlSGjlypLF/f39/ubu7a8GCBQoMDFSlSpVkZ2enAQMGqHnz5jk+bgAAAAAAALkp34yEkqTvv/9ezzzzjC5evKg5c+ZoxowZ2rBhw0P16e7uroiICE2aNEm1atVSZGSkFixYoL1796pSpUr67LPPTEKdrKhSpYq2bNmiunXravXq1dq+fbv8/f0VFhamgQMHmq3v7++vffv26Y033lCxYsW0YcMGLVmyRHFxcQoICFBoaKjatWtnXH/y5Mnq2bOnbt68qZUrV2rVqlWyt7fX8OHDtXfvXlWtWtW4rrOzs3799VcFBQUpKipKoaGhmjFjhg4fPpzzgwYAAAAAAJDLDOnp6em2LgL5i6+vryQpOjraxpUAAID8IDEpSUPeHSdJClsfqdR7phcAkD+c3bXE1iUAwH3lq5FQAAAAAAAA+GcihAIAAAAAAIDVEUIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6e1sXAAAAgILj2OZ5cnJ0tHUZAACgAGIkFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqyOEAgAAAAAAgNURQgEAAAAAAMDqCKEAAAAAAABgdfa2LgAAAAAFh0+LnkpNS7N1GQDucnbXEluXAABZwkgoAAAAAAAAWB0hFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqyOEAgAAAAAAgNX960Mog8Egb29vW5eRK/r16yeDwaDw8HBblwIAAAAAAGDiXx9CAQAAAAAAwPoIoQAAAAAAAGB1hFAAAAAAAACwOkKo+1ixYoWCgoJUsmRJOTs7q3r16ho1apSuXLlist6QIUNkMBg0derUTPtq2LChDAaD9u7da9IeHx+vV155RZUrV5azs7NKlSqlTp06KSIiItO+fvjhB/n5+cnFxUWenp7q16+fzpw581DPFQAAAAAAwJoIoTIxduxYdezYUeHh4WrYsKFCQkJ08+ZNjRs3Tk2bNtXZs2eN6/bp00eSNGfOHIt9HTx4ULt371adOnVUt25dY/uWLVtUr149ff3113JwcFDHjh1Vu3ZtrV69Wq1bt9bPP/9s1teoUaP0/PPPa//+/WrdurVat26tlStXqmnTprp06VIuHwUAAAAAAIDcYW/rAvKjHTt26N1335Wrq6t+//13NW3aVJKUmJioZ555Rr/88osGDx6sBQsWSJKaNm2qKlWqaPPmzTp+/LgeeeQRk/5++uknSX+HVZKUkJCgbt26KSEhQbNnzzZZtnPnTrVv314DBw5U27ZtVaZMGUnS1q1bNX78eBUvXlzr169X/fr1JUnXr1/XE088obCwMOsdFAAAAAAAgIfASCgLpkyZorS0NL366qvGAEqSnJycNGXKFLm4uGjx4sWKj483LuvTp4/S09MtjoaaM2eODAaDevfubWz74YcfdPr0aQ0bNswkgJKkRo0aafTo0bp+/bpmz55tbJ86darS09M1dOhQYwAlSa6urpo8ebIMBkO2nqevr6/Fn5iYmGz1AwAAAAAA8CCEUBZs2rRJkszCIUlyd3dX+/btlZaWps2bNxvbM7skb8uWLfrrr7/UunVreXl5GdvXrFkjSeratavFGlq1aiVJ2r59u1ldPXv2NFu/Vq1aqlev3oOfHAAAAAAAgA1wOZ4Fp06dkiR5e3tbXJ7RfvLkSWNb1apV1bhxY+3YsUP79u1TnTp1JFm+FE+SYmNjJUktWrS4by0XLlwwq6tixYqZ1hUVFXXf/u4WHR1tsd3X1zfLfQAAAAAAAGQFIVQOZHbZW9++fbVjxw7NmTNHY8eOVUpKiubPny8nJyd1797dZN20tDRJUvfu3VWkSJFM91WjRo3cKxwAAAAAAMBGCKEsKFeunI4dO6a4uDjVqlXLbHnGKKby5cubtPfo0UPDhw/X3Llz9cknn2jNmjU6f/68nnzySZUsWdJk3QoVKujQoUMaNWqUGjZsmKW6ypYtq9jYWMXFxalmzZpmy+Pi4rL4DAEAAAAAAPIWc0JZkDEf09y5c82WnT9/XqtXr5bBYDC7lM7Dw0Pt2rVTXFycNm/enOmleJIUFBQkSVq8eHG265o/f77ZsoMHD2brUjwAAAAAAIC8RAhlweDBg2VnZ6dJkyZp586dxvakpCS9+uqrunXrlrp27Woy0XiGvn37SpKmTZumpUuXqnjx4urUqZPZeoMGDZK7u7vGjx+vadOmGS/Py5CSkqLVq1frzz//NLa99NJLkqSJEydqz549xvYbN27o1VdfVXp6+sM9cQAAAAAAACshhLKgSZMm+uijj5SQkKBmzZopKChIvXr1UpUqVfTzzz+ratWq+vrrry1uGxISosKFC2vWrFm6ceOGunfvLicnJ7P1SpQoYQypBg0aJG9vbz3++OPq06ePHn30UZUpU0bBwcE6evSocZvmzZtr5MiRunLliho3bqzg4GD16NFDlStX1uHDh9W5c2erHRMAAAAAAICHQQiVibffflvLly9XQECAduzYoUWLFsnJyUlvvPGGtm3bJg8PD4vbubq66oknnjA+tnQpXgZ/f3/t27dPb7zxhooVK6YNGzZoyZIliouLU0BAgEJDQ9WuXTuTbT777DNNnz5dNWvWVHh4uMLDwxUUFKQtW7aoVKlSufPkAQAAAAAAcpkhnWu4cA9fX19JUnR0tI0rAQAA+UFiUpKGvDtOkhS2PlKp90wjAMC2zu5aYusSACBLGAkFAAAAAAAAqyOEAgAAAAAAgNURQgEAAAAAAMDqCKEAAAAAAABgdYRQAAAAAAAAsDp7WxcAAACAguPY5nlycnS0dRkAAKAAYiQUAAAAAAAArI4QCgAAAAAAAFZHCAUAAAAAAACrI4QCAAAAAACA1RFCAQAAAAAAwOoIoQAAAAAAAGB19rYuAADytaoGW1cAALZXyEHqMtrWVQAAgAKOkVAAAAAAAACwOkIoAAAAAAAAWB0hFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAqyOEyiWxsbEyGAwKDAw0aQ8NDZXBYNCYMWOy1Z/BYJC3t3eu1QcAAAAAAGBLBTKEyizwAQAAAAAAQP5kb+sC/inKly+vAwcOqHDhwrYuBQAAAAAAIN8hhMolDg4OqlGjhq3LAAAAAAAAyJcK3OV4Y8aMkY+PjyRpw4YNMhgMxp9+/foZ17t06ZLeeust1apVSy4uLipevLjatm2r5cuXm/V59+V9CQkJGj58uHx8fOTg4KBhw4ZJkry9vWUwGJSenq7JkyerXr16Kly4sPz8/Mz6yMyhQ4fUrVs3ubm5qUiRImrRooVWrFiR7WNw4MAB9evXT15eXnJycpKHh4d69uyp6OjobPcFAAAAAACQFwrcSCg/Pz9169ZNCxculIeHh4KDg43LWrZsKUk6fPiw2rVrp/j4eHl7e6tDhw66du2atm7dqs6dO+uzzz7TyJEjzfq+deuWAgICFBcXp4CAADVo0EAlS5Y0Weell17Sjz/+qICAANWsWVNJSUlZqjsmJkZNmzZVqVKl1L59e506dUqbNm1Sp06dNGPGDPXv3z9L/SxZskQ9e/ZUYmKi/Pz85O/vr/j4eM2fP19hYWFauXKlWrdunaW+AAAAAAAA8kqBC6FCQkLk5+enhQsXqkaNGgoNDTVZnpqaqu7duys+Pl7jx4/XiBEjZGd3Z8DX0aNH1b59e40aNUrBwcGqXbu2ybbbt29Xs2bN9Ndff6lEiRIW979o0SJFRkbK19c3W3XPnj1bzz77rGbMmCF7+zuHffny5QoJCdErr7yi9u3bq3z58vftIzY2Vn379pWDg4OWL1+udu3aGZetWrVKXbp0Ud++fXX06FE5Ojpmqz4AAAAAAABrKnCX4z1IWFiY9u3bp27duun11183BlCSVKVKFX3xxRdKTU3V9OnTLW4/adKkTAMoSXrzzTezHUBJkqurqyZOnGgMoCSpU6dO6t69u27evKkff/zxgX1MnDhRN27c0NixY00CKEkKDg7Wyy+/rPj4eP36669ZqsnX19fiT0xMTPaeHAAAAAAAwAP840KoNWvWSJK6du1qcXmrVq0k3Rn1dK+yZcuqUaNG9+2/S5cuOaqrffv2Zpf2SVKvXr0kSZs2bXpgHw/z3AAAAAAAAGypwF2O9yCxsbGSpD59+qhPnz6ZrnfhwgWztkceeeSB/WdlHUsqVqxosd3b21uSdOrUqQf2kfHcHnTZnqXnZklmE5nnZKQXAAAAAADA/fzjQqi0tDRJdy5P8/DwyHS90qVLm7U5Ozs/sP+srGMtGc/tueeeu+96TZs2zYtyAAAAAAAAsuwfF0JVqFBBkjRw4EB169bNxtX8LS4u7r7t5cqVe2AfFSpUUExMjL744gu5ubnlan0AAAAAAADWVCDnhMq481tKSorZsqCgIEnS4sWL87SmB1mzZo2uXLli1j5v3jxJUsuWLR/YR359bgAAAAAAAA9SIEOo0qVLy8HBQTExMUpNTTVZ1q1bN9WqVUs//fSTPvroIyUmJposT09P1+bNm7V58+a8LFnXr1/X8OHDTYKzlStXav78+XJxcVH//v0f2MeIESPk4uKikSNHatGiRWbLExMTtWDBAp04cSJXawcAAAAAAHhYBfJyPEdHRwUHByssLEz16tVTgwYN5OjoqBYtWqh///5asmSJOnTooPfee09TpkxR3bp15e7urgsXLigqKkrnzp3ThAkT1KJFizyruU+fPlq0aJHCw8PVtGlTnT59Whs3blR6eromTZpkvIzwfqpUqaK5c+eqd+/e6tatm6pUqaKaNWuqSJEiOnnypHbv3q0bN24oMjIyS/0BAAAAAADklQIZQknS999/r5EjR+q3337TnDlzlJqaqpSUFPXv319Vq1ZVZGSkpkyZokWLFmnr1q1KSUmRp6en6tevry5duujpp5/O03qrVKmiLVu26K233tLq1at1+/Zt+fv76+2331anTp2y3M8TTzyhvXv36ssvv9Rvv/2m3377TQ4ODipXrpw6d+6srl27qlatWlZ8JgAAAAAAANlnSE9PT7d1EchffH19JUnR0dE2rgTIB6oabF0BANhcYiEHDekyWpI06b9vyun/5+cEAADIjgI5JxQAAAAAAAAKFkIoAAAAAAAAWB0hFAAAAAAAAKyOEAoAAAAAAABWRwgFAAAAAAAAq7O3dQEAkK8d4QaiAKCkJOndcbauAgAAFHCMhAIAAAAAAIDVEUIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6QigAAAAAAABYHSEUAAAAAAAArM7e1gUAQL5U1WDrCgAg/yjkIHUZbesqAABAAcdIKAAAAAAAAFgdIRQAAAAAAACsjhAKAAAAAAAAVkcIBQAAAAAAAKsjhAIAAAAAAIDVEUIBAAAAAADA6gih8pHQ0FAZDAaNGTPG1qUAAAAAAADkKkIoAAAAAAAAWB0hFAAAAAAAAKzuXxNC3b59W87OzvL29jZbFhISIoPBoJYtW5ota9Sokezs7HT+/HlJ0qZNm/TKK6+obt26KlmypFxcXFSjRg2NGjVKV65csbjviIgIhYSEqGLFinJycpKnp6eaNGmiUaNG6fr165KkwMBA9e/fX5L0wQcfyGAwGH9CQ0NN+jtw4ID69esnLy8vOTk5ycPDQz179lR0dHTODxAAAAAAAIAVGdLT09NtXUReCQgI0MaNG3Xs2DFjGJWWlqbSpUvr8uXLcnR01OXLl1W4cGFJ0tWrV1WqVCnVrFlTf/75pyTJ399fe/bsUd26deXl5aXbt29r9+7dOn36tHx9fbV161a5uroa9xkWFqaQkBClp6erSZMm8vHx0ZUrV3TkyBHFxMQYa/n000+1fPlybd68WfXq1ZOfn5+xj4EDBxoDsiVLlqhnz55KTEyUn5+fqlSpovj4eG3fvl0uLi5auXKlWrdu/VDHydfXV5IItfDvVtVg6woAIN9ILOSgIV1GS5Im/fdNOTk62rgiAABQENnbuoC8FBgYqI0bNyo8PFz9+vWTJO3Zs0eXL1+Wr6+voqOjFRERoXbt2kmSNm7cqLS0NAUGBhr7eP/999W8eXMVL17c2JaYmKghQ4Zo2rRp+vLLL/Xee+8Zl33++edKS0vTggUL1K1bN5N6duzYITc3N0nSqFGj5Onpqc2bNyskJMTi5OSxsbHq27evHBwctHz5cmOdkrRq1Sp16dJFffv21dGjR+WYhT8OM8Kme8XExKhy5coP3B4AAAAAACCr/jWX40kyhknh4eHGtox/ZwRHlpYFBAQY2x577DGTAEqSnJycNHHiRNnb22vp0qUmyzIu47s7MMrQuHFjFS1aNMv1T5w4UTdu3NDYsWPN+gsODtbLL7+s+Ph4/frrr1nuEwAAAAAAIC/8q0ZC+fv7y8nJySxoKlq0qLp166aKFStaDKHuHgklSSdPnlRYWJgOHjyohIQEpaWlSZIcHR115MgRk3UbNmyoAwcO6JlnntHo0aPVsGFD2dnlLPtbs2aNJKlr164Wl7dq1UqTJk3S9u3b9eSTTz6wv8wut8tshBQAAAAAAEBO/atCKBcXFzVp0kSbNm1SbGysHnnkEW3atEmtWrVSoUKFFBgYqLlz5+rmzZtKTk5WVFSUatWqpTJlyhj7+PLLLzVq1CglJydnaZ+ffPKJ9u3bp7CwMIWFhalkyZJq2bKl8dI5Z2fnLNcfGxsrSSpfvvx917tw4UKW+wQAAAAAAMgL/6oQSrozqmnTpk0KDw9XvXr1dPnyZeNIp8DAQM2cOVMRERG6deuW0tLSTC7F27p1q0aMGKHixYvrq6++UmBgoDw9PeXk5CRJKleunE6fPm2yPy8vL+3cuVPr1q3T8uXLtWHDBmMgNX78eG3ZssU4L9SDZIy4eu655+67XtOmTbN6OAAAAAAAAPLEvy6ECggI0EcffaTw8HBdvnxZkkxCKOnOZXi3bt0yaZOkxYsXS5I+/vhjsyDo1q1bOnPmjMV92tvbq3379mrfvr0kKS4uTgMGDNC6des0btw4jR8/Pku1V6hQQTExMfriiy+yHFwBAAAAAADkB/+qicklqXnz5nJ0dFR4eLjCw8NVrFgxNWjQQJLk7e1tnBfK0qTkGaFVhQoVzPr95ZdflJ6enqUaKlasqDfffFOS9OeffxrbM+5ol5KSYnG7oKAgSX+HYQAAAAAAAAXFvy6EypgXKi4uTmvWrDHOB5UhMDBQ27dvV1RUlGrUqCEPDw/jsmrVqkmSZsyYYTIn1P79+42h0r0mTJhgcYTUihUrJN25XC9DuXLlJEmHDh2y2NeIESPk4uKikSNHatGiRWbLExMTtWDBAp04cSLT5w8AAAAAAGAL/7oQSvp7dNPt27fN7nwXGBio5ORkpaWlmS3r37+/PD09FRYWpurVq6tHjx4KCgqSn5+fWrVqpYoVK5rt64MPPlD58uXVoEED9ejRQ08//bSqV6+ur776SqVKldLIkSON6/r7+8vd3V0LFixQYGCgBgwYoIEDByoiIkKSVKVKFc2dO1fJycnq1q2bqlatqi5duqhXr15q3bq13Nzc9NRTTzExOQAAAAAAyHf+lSHU3eGSpRAqs2Vubm7asWOHevfuraSkJC1btkwnT57URx99pLlz51rc1+TJk9WzZ0/dvHlTK1eu1KpVq2Rvb6/hw4dr7969qlq1qnFdZ2dn/frrrwoKClJUVJRCQ0M1Y8YMHT582LjOE088ob179+o///mPDAaDfvvtN/366686d+6cOnfurPnz56tWrVo5PjYAAAAAAADWYEjP6kRG+Nfw9fWVJEVHR9u4EsCGqhpsXQEA5BuJhRw0pMtoSdKk/74pp/+fxxIAACA7/pUjoQAAAAAAAJC3CKEAAAAAAABgdYRQAAAAAAAAsDpCKAAAAAAAAFidva0LAIB86Qj3bAAAo6Qk6d1xtq4CAAAUcIyEAgAAAAAAgNURQgEAAAAAAMDqCKEAAAAAAABgdYRQAAAAAAAAsDpCKAAAAAAAAFgdIRQAAAAAAACszt7WBQCARVUNtq4AAJChkIPUZbStqwAAAAUcI6EAAAAAAABgdYRQAAAAAAAAsDpCKAAAAAAAAFgdIRQAAAAAAACsjhAKAAAAAAAAVkcIBQAAAAAAAKsjhAIAAAAAAIDVEUIBAAAAAADA6gihAAAAAAAAYHWEUAAAAAAAALA6Qqgsio2NlcFgUGBgoBISEjR06FB5eXnJ2dlZNWvW1IQJE5SWlma2XXx8vAYNGqSKFSvKyclJ7u7u6tq1q3bs2JHpvrZs2aInnnhCZcqUkZOTk7y9vfWf//xHp06dMls3NDRUBoNBY8aM0eHDh9WzZ095eHjIzs5OS5Ysyc1DAAAAAAAAkGOEUNmUmJiotm3b6n//+5+aNGmioKAgxcXFafjw4RowYIDJuvv27VODBg00bdo0ubi4qGvXrqpataoWL16s5s2b65dffjHrf/bs2WrVqpWWLVum6tWrq2vXrnJyctLUqVPVoEEDHTx40GJdhw4dUuPGjbV9+3a1adNGQUFBcnBwsMoxAAAAAAAAyC57WxdQ0GzdulV169bVkSNHVLp0aUlSTEyMWrdurZkzZyokJEQhISFKT09Xnz59dOHCBb3xxhv69NNPZTAYJEkLFy7U008/rQEDBqhly5YqW7aspDujpl588UVJ0tKlS9WlSxdJUlpamkaMGKGJEyfqmWeesTiKat68eXrllVc0ceJEFSpUKEvPxdfX12J7TEyMKleunL0DAwAAgP9r787Doqr+OI5/hh3FXUDcwH1BTXHFDbQ0KxeUslwqNUvNss2sX6mZ2W5u2WKrllu5YJrmhhEI4r6SqSmg4o64szO/P4yJcYYEdRyX9+t5eB4559xzv/fO3DvOl3POBQAA/4GRUNdg/PjxpgSUJFWrVk2jRo2SJE2dOlWSFBERoZ07d6py5coaN26cKQElSaGhoQoJCdGFCxf03Xffmcq/+eYbpaamqmfPnqYElCQ5ODjogw8+UPny5bVp0yZFR0dbxOTp6akPP/ywwAkoAAAAAACAm4kkVCGVLl1aHTp0sCjv1auXJCkmJkY5OTmKioqSJPXs2dPqtLjHH39ckkzt8v67T58+Fu1dXV31yCOPWGyT67777lORIkUKdSxxcXFWfxgFBQAAAAAAbjSSUIXk6+trtbxEiRIqWbKkUlNTlZKSYlpE3M/Pz2r73PKkpCRT2bVsk6ty5coFiB4AAAAAAMA+SELZSd7peTdiGzc3t+sJBwAAAAAAwKZIQhXSwYMHrZafO3dOZ86ckbu7u0qWLKny5ctLkhITE622T0hIkCRVqFDBVHYt2wAAAAAAANwOSEIVUnJyssLDwy3K586dK0kKDAyUo6Oj2rRpI0maN2+esrOzLdrPnDlTkkzt8v57zpw5Fu0zMjI0b948i20AAAAAAABuByShrsHw4cOVnJxs+j0+Pl5jx46VJA0dOlSSFBwcrPr16yshIUGjR4+W0Wg0tQ8LC9PChQvl4eGhAQMGmMqfeuopubu7a+7cuVq6dKmpPCcnR2+88YaSkpLUuHFjtWrVytaHCAAAAAAAcEM52TuA202LFi2UkZGh6tWrq3379srMzFR4eLguXbqkvn37qkePHpIur980a9YstWvXTu+9957CwsLUsGFDHTx4UNHR0XJyctK3334rHx8fU9+VK1fWtGnT1K9fP3Xp0kWtWrVSpUqVtGXLFu3Zs0fe3t6mEVQAAAAAAAC3E0ZCFZKrq6vWrFmj3r17KzY2VitWrFClSpU0fvx4TZ8+3axt/fr1tWXLFj399NO6cOGC5s+frz179igkJETR0dHq2bOnRf+PP/64oqKi1LlzZ+3evVvz589XamqqhgwZos2bN6t27do36UgBAAAAAABuHIMx7zwx5CshIUFVqlRRUFCQIiIi7B2OTfn7+0uS4uLi7BwJ7mo1Cv8ESQCAbaQ7OmtY11GSpCnjXpOri4udIwIAALcjRkIBAAAAAADA5khCAQAAAAAAwOZIQgEAAAAAAMDmeDpeAfn5+YnlswAAAAAAAK4NSSgAt6Z9JH0B4JaRkSGN/NDeUQAAgNsc0/EAAAAAAABgcyShAAAAAAAAYHMkoQAAAAAAAGBzJKEAAAAAAABgcyShAAAAAAAAYHMkoQAAAAAAAGBzTvYOAMAdrobB3hEAAK6Xo7PUdZS9owAAALc5RkIBAAAAAADA5khCAQAAAAAAwOZIQgEAAAAAAMDmSEIBAAAAAADA5khCAQAAAAAAwObsmoQyGAzy8/OzZwgAAAAAAAC4CWyWhEpISJDBYFBwcLCtdoF/kMwDAAAAAAC3OqbjAQAAAAAAwOZIQgEAAAAAAMDmbJKEGjNmjKpUqSJJ+uOPP2QwGEw//fr1s2ifnZ2tDz/8UDVr1pSrq6sqVaqk1157Tenp6Vb7v3Tpkt5//301atRIHh4e8vDwUIsWLTRjxowCxxgREWE1npycHJUuXVoGg0EjR440qzt79qycnJxUr149i/6WLVumDh06qFSpUnJzc1OtWrX0+uuv68yZMxZtx4wZI4PBoOnTp2vDhg3q3LmzypQpI4PBoG3btkmSTp48qddff11169aVh4eHSpQooZo1a+qJJ57Qhg0bJEnTp0+XwWCQJCUmJpqdZ6ZBAgAAAACAW4mTLTpt2LChQkNDtWDBAnl7e6tTp06mutatW1u07927t5YtW6bg4GDVqlVLUVFR+uijj5SUlKSZM2eatT1x4oQ6dOigHTt2qFy5cgoKCpLRaFRMTIz69eunTZs26dNPP71qjC1atJCrq6siIiLMyrdv366UlBRJsqiLjIxUdna2RYLn/fff1xtvvCEnJycFBQWpbNmyio6O1ocffqiwsDBFRkbK29vbIobIyEg988wzqlmzpjp27KgjR47IwcFB58+fV/PmzRUfH69KlSqpQ4cOcnJy0sGDBzV37lxVrVpVzZo1U/Xq1fXkk09qxowZKlq0qB5++GFT37Vr177qOQAAAAAAALhZbJKECgkJUcOGDbVgwQLVrl1b06dPz7dtYmKiihQpon379qlcuXKSpPj4eAUEBGjWrFl6++23Va1aNVP7/v37a8eOHXrhhRf04YcfytXVVZJ0/Phxde7cWVOnTtVDDz1klviyxs3NTc2bN1dkZKQSEhJMC3vnJp78/f21ceNGXbp0SUWKFDGry5uE2rhxo0aOHCkPDw+tXr1azZs3lySlp6fr8ccf17x58zR06FDNnz/fIobvv/9eH374oUaMGGFRHh8fr65duyosLEwODv8OWDt58qSOHz8u6XJCr3Xr1poxY4bKli37n+cZAAAAAADAnm6JNaGmTJliSkBJUpUqVdS3b19JUlRUlKl827ZtWrZsmZo2baoJEyaYElCS5O3tra+++kqS9MUXXxRov7nJpLwjniIiIlSsWDG99NJLysjIUExMjFmdJAUFBZnKpk6dqpycHD3//POmBJQkubq6aurUqXJ3d1dYWJgOHTpksf/69evr1VdftSg/efKkJKl9+/ZmCShJ8vT0tDod8Fr4+/tb/dm/f/8N6R8AAAAAACCX3ZNQzs7OateunUV5zZo1JUlHjx41la1cuVLS5ZFWVyZnJJnWiMpdM+lqcpNJucmlnJwcRUVFqXXr1rr33nvN6s6ePatt27apbt268vT0NPWRmyTr06ePRf9eXl7q2LGjcnJyFB0dbVHfuXNn05pOeTVu3FiS9PHHH2vu3Lk6f/58gY4HAAAAAADgVmWT6XiFUa5cOTk6OlqUFytWTJLMFidPSEiQJL355pt688038+0zLS2tQPsODAw0Wxcqdz2odu3ayc/PT76+vqa6yMhI5eTkWKwHdeTIEUkyTee7Um55UlKSRV3lypWtbnPvvffqpZde0qRJk9SrVy85OTkpICBAHTp00IABA1S1atUCHd/VxMXFWS339/e/If0DAAAAAADksnsSytqIpvzk5ORIurwWUt51oq6Vu7u7mjVrpqioKCUkJFis+RQcHKw5c+bo0qVLVqfiFYS1kU653Nzc8q2bMGGCBg0apF9++UWrV69WdHS0NmzYoI8++khz5sxRaGhooeIAAAAAAACwJ7snoQqjYsWKki5Px3vllVduSJ9BQUGKiopSRESEIiIiVLx4cQUEBEi6nISaMWOGYmJirC5KLknly5dXfHy8EhMTVbduXYv+c0dvVahQodCx1apVSyNGjNCIESOUlpamqVOn6tVXX9WQIUNIQgEAAAAAgNuKzdaEcnFxkSRlZWXdsD47dOggSQoLC7thfeYmldasWaOoqCi1adPGND0wt+6XX37Rtm3bVKdOHXl5eZlt36ZNG0nSnDlzLPo+efKkVqxYIYPBoFatWl1XnG5ubho+fLh8fHx08uRJnThxwlTn7Ox8Q88zAAAAAADAjWazJFTZsmXl7Oys/fv3Kzs7+4b02bx5c3Xo0EHR0dEaOnSozp07Z9Fm+/btWr58eYH7bNmypVxcXDRv3jylpKSYjXTKXRfqm2++UU5OjtWpeEOHDpWDg4OmTJmiTZs2mcozMjL0/PPPKzU1VT169FClSpUKHNOiRYsUGxtrUb5582YdP35cHh4eKlmypKm8fPnyOn78uM6cOVPgfQAAAAAAANxMNh0J1alTJx07dkz33HOPnnjiCQ0cOFDff//9dfU7c+ZMNWrUSJ9//rl8fX3Vrl079enTR507d1blypXVsGHDQiWh3N3d1bRpU9Ni5ldOtwsODs63TpKaNWumd955R+fOnVNgYKA6dOigXr16qXr16vrpp59Uo0YNffbZZ4U6xoiICAUGBqpixYrq0qWL+vTpo3bt2ql58+bKycnR22+/bRppJkldu3ZVVlaWAgIC1LdvXw0cOFAff/xxofYJAAAAAABgSzZdE+qbb77R8OHDtWrVKs2ePVvZ2dnKyspS//79r7lPLy8vxcTE6Ouvv9bcuXO1detWxcTEyNvbW1WrVtWwYcP02GOPFarP4OBgRUdHq0SJEmrUqJFF3YwZMyTlvyj5G2+8oXvuuUcTJ07Uxo0blZqaqsqVK2vEiBF6/fXXVapUqULF069fPzk5OSkyMlIbNmzQ2bNnVa5cOT344IN64YUXdO+995q1f//992U0GvXLL7/op59+UlZWloKCgvTqq68War8AAAAAAAC2YjAajUZ7B4Fbi7+/vyQpLi7OzpHgjlAj/ydEAgBuD+mOzhrWdZQkacq41+SaZ0Q2AABAQdlsOh4AAAAAAACQiyQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbM7J3gEAuMPt4wGcAHDby8iQRn5o7ygAAMBtjpFQAAAAAAAAsDmSUAAAAAAAALA5klAAAAAAAACwOZJQAAAAAAAAsDmSUAAAAAAAALA5klAAAAAAAACwOSd7BwDgBqhhsHcEAIA7maOz1HWUvaMAAAC3OUZCAQAAAAAAwOZIQgEAAAAAAMDmSEIBAAAAAADA5khCAQAAAAAAwOZIQgEAAAAAAMDm7tgkVEJCggwGg4KDg+0ax/Tp02UwGDRmzBi7xgEAAAAAAGBPd2wSCgAAAAAAALcOJ3sHcKfr3r27WrRoobJly9o7FAAAAAAAALshCWVjJUqUUIkSJewdBgAAAAAAgF3dFdPxzp07pxdeeEGVKlWSm5ub6tSpo4kTJyonJ8ei7aFDhzRo0CD5+vrK1dVVXl5e6tGjhzZu3GjRNu+6U+fOndPLL7+sKlWqyNnZWS+++KKk/NeE6tevnwwGgyIiIhQZGan27durWLFiKl68uB566CH9+eefVo/l4sWLev311+Xn5yc3NzdVr15d77zzjjIzM+Xn5yeDwXDd5wsAAAAAAOBGu+NHQqWnp6t9+/bav3+/2rdvr4yMDIWHh+vll1/W9u3bNX36dFPbnTt3qn379jp16pRq1aqlHj166ODBgwoLC9OSJUs0e/ZsPfLIIxb7SE1NVVBQkBITExUUFKSAgACVKlWqQPEtWbJEkydPVpMmTfTggw9q27ZtWrZsmdavX69du3apXLlyZsdy3333KTY2VqVLl1bnzp2Vnp6uDz74QFu2bLnucwUAAAAAAGArd3wSKjY2Vg0aNNC+fftM6zLt379fbdu21YwZMxQSEqKQkBAZjUb16dNHp06d0ogRI/TBBx+YRhUtWLBAPXv21IABA9S6dWv5+PiY7WPDhg0KDAzUgQMHVLJkyULFN2nSJC1YsEAhISGSpOzsbD366KNasGCBPv/8c40dO9bU9uOPP1ZsbKyaNWumFStWmPaVkJCgtm3b6tChQ9d2kgAAAAAAAGzsrpiON378eLOFwatVq6ZRo0ZJkqZOnSpJioiI0M6dO1W5cmWNGzfObFpbaGioQkJCdOHCBX333XdW9zFlypRCJ6AkqVevXqYElCQ5Ojrqf//7nyQpMjLSrO2XX34pSfrkk0/M9uXn56fRo0cXet/+/v5Wf/bv31/ovgAAAAAAAP7LHZ+EKl26tDp06GBR3qtXL0lSTEyMcnJyFBUVJUnq2bOnnJ2dLdo//vjjkmRql5ePj4+aNGlyTfF17NjRoqxmzZqSpKNHj5rKEhMTlZSUpHLlyql169YW2zz66KPXtH8AAAAAAICb4Y6fjufr62u1vESJEipZsqTOnDmjlJQUHTlyRNLlUUXW5JYnJSVZ1FWuXPma46tYsaJFWbFixSRdXgMqV25CqlKlSlb7KVasmOl4CiouLs5qub+/f4H7AAAAAAAAKIg7fiTUjfJfT51zc3O75n4dHHgJAAAAAADAne+Oz4AcPHjQavm5c+d05swZubu7q2TJkipfvryky9PerElISJAkVahQwSZxXk3uYuj5LT5+/vz5Qo2CAgAAAAAAuJnu+CRUcnKywsPDLcrnzp0rSQoMDJSjo6PatGkjSZo3b56ys7Mt2s+cOVOSTO1uNl9fX1WoUEHHjh1TTEyMRf28efPsEBUAAAAAAEDB3PFJKEkaPny4kpOTTb/Hx8dr7NixkqShQ4dKkoKDg1W/fn0lJCRo9OjRMhqNpvZhYWFauHChPDw8NGDAgJsbfB6DBw+WJL3yyis6e/asqTwxMdF0PAAAAAAAALeiO35h8hYtWigjI0PVq1dX+/btlZmZqfDwcF26dEl9+/ZVjx49JF1e82nWrFlq166d3nvvPYWFhalhw4Y6ePCgoqOj5eTkpG+//dY0Lc4eXn31VS1dulSxsbGqVq2a2rVrp/T0dK1Zs0b33nuvjEaj2RP1AAAAAAAAbhV3/EgoV1dXrVmzRr1791ZsbKxWrFihSpUqafz48Zo+fbpZ2/r162vLli16+umndeHCBc2fP1979uxRSEiIoqOj1bNnT/scxD9cXV21atUqjRgxQkWLFtXixYsVFxenV155RT/99JOOHz+uMmXK2DVGAAAAAAAAawzGvPPOcNuKjY1VYGCgOnXqpN9+++26+vL395ckxcXF3YjQcDPUyP/pjQAAXK90R2cN6zpKkjRl3GtydXGxc0QAAOB2dMePhLrTbN26VTk5OWZlBw4c0KBBgyRJffv2tUdYAAAAAAAA/+mOXxPqTtOrVy+dPXtW9evXV5kyZXTw4EFt3rxZ6enp6tq1q3r37m3vEAEAAAAAACyQhLrNPP/88/r555+1Y8cOnT59Wm5ubmrYsKH69u2rwYMHy2BgWhYAAAAAALj1kIS6zQwdOlRDhw61dxgAAAAAAACFwppQAAAAAAAAsDlGQgF3gn085BIAYEMZGdLID+0dBQAAuM0xEgoAAAAAAAA2RxIKAAAAAAAANkcSCgAAAAAAADZHEgoAAAAAAAA2RxIKAAAAAAAANkcSCgAAAAAAADbnZO8AgDtGDYO9IwAAwDYcnaWuo+wdBQAAuM0xEgoAAAAAAAA2RxIKAAAAAAAANkcSCgAAAAAAADZHEgoAAAAAAAA2RxIKAAAAAAAANkcSyk6ys7M1evRoVatWTS4uLjIYDOrXr98199evXz8ZDAZFRETcsBgBAAAAAABuFCd7B3C3mjx5st555x2VL19ePXr0kJubm1q3bm3vsAAAAAAAAGyCJJSdLFq0SJIUFRWlqlWr2jcYAAAAAAAAG2M6np0cPnxYkkhAAQAAAACAuwJJKElLly7VgAEDVKdOHRUvXlxFixbVPffco/fee0/p6emmdhEREVbXbsrJyVHp0qVlMBg0cuRIs7qzZ8/KyclJ9erVk/Tv2k3x8fGSJIPBYPpJSEgwlfn5+VmNdfr06TIYDBozZswNOXYAAAAAAICbgel4kp566imlpqaqXr16atCggc6ePasNGzbozTffVHh4uFauXClHR0e1aNFCrq6uFot/b9++XSkpKZJkURcZGans7GwFBwdLkmndp/nz5+vixYt68sknTW09PDxsdowAAAAAAAD2RBJK0rRp09SxY0e5u7ubys6fP6/evXvr119/1axZs/TEE0/Izc1NzZs3V2RkpBISEkyjlXITT/7+/tq4caMuXbqkIkWKmNXlJqEGDhyogQMHKiIiQhcvXtT06dNv0lECAAAAAADYD9PxJHXr1s0sASVJxYoV08SJEyVJv/zyi6k8N5mUd8RTRESEihUrppdeekkZGRmKiYkxq5OkoKAg2wR/Hfz9/a3+7N+/396hAQAAAACAOwwjof6xb98+LVu2TH///bcuXryonJwcGY1GU12u3GRSRESE+vXrp5ycHEVFRal169a69957TXX33Xefzp49q23btqlu3bry9PS8+QcFAAAAAABwi7jrk1BGo1HDhw/XxIkTTUmnK50/f97078DAQLN1oXLXg2rXrp38/Pzk6+trqouMjFROTo5p9NStJi4uzmq5v7//TY4EAAAAAADc6e766Xg//fSTJkyYoIoVK2r+/PlKSkpSRkaGjEaj6cl4eZNT7u7uatasmRITE5WQkGCx5lNwcLBpXShbTMXLycm5YX0BAAAAAADcLHd9EiosLEyS9MUXXyg0NFTly5eXs7OzJOnAgQNWt8k7JS8iIkLFixdXQECApMtJqNx1oa5MUBWUs7OzLly4YLXu0KFDheoLAAAAAADgVnDXJ6FSUlIkSRUrVrSo+/nnn61uk5tUWrNmjaKiotSmTRs5Ojqa1f3yyy/atm2b6tSpIy8vr0LF5OPjo+TkZCUnJ1vUrV69ulB9AQAAAAAA3Aru+iRUzZo1JUlfffWV2bS7qKgoffzxx1a3admypVxcXDRv3jylpKSYjXTKXRfqm2++UU5OzjVNxcvdZty4cWblH330kdauXVvo/gAAAAAAAOztrk9CDRs2TEWLFtXnn3+uevXqqVevXmrbtq2CgoI0ePBgq9u4u7uradOmSktLk2Q53S44ODjfuoJ47bXX5O7urkmTJqlRo0Z6+OGHVatWLY0ZM0bPPvtsofsDAAAAAACwt7s+CVWzZk1t2rRJXbp00alTp7R48WJduHBB06ZNy3cklPRvcqlEiRJq1KiR1Trp2hYl9/f315o1axQcHKy9e/dq1apVqlatmtatW6emTZsWuj8AAAAAAAB7MxjzzkEDdDkJJklxcXF2juQ2U8Ng7wgAALCJdEdnDes6SpI0ZdxrcnVxsXNEAADgdnTXj4QCAAAAAACA7ZGEAgAAAAAAgM2RhAIAAAAAAIDNkYQCAAAAAACAzZGEAgAAAAAAgM052TsA4I6xjwdNAgDuUBkZ0sgP7R0FAAC4zTESCgAAAAAAADZHEgoAAAAAAAA2RxIKAAAAAAAANkcSCgAAAAAAADZHEgoAAAAAAAA2RxIKAAAAAAAANudk7wCAm6KGwd4RAABw+3J0lrqOsncUAADgNsdIKAAAAAAAANgcSSgAAAAAAADYHEkoAAAAAAAA2BxJKAAAAAAAANgcSSgAAAAAAADYHEmoGyQhIUEGg0HBwcF2i2HMmDEyGAyaPn263WIAAAAAAACwhiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiSUDZw7d04vvPCCKlWqJDc3N9WpU0cTJ05UTk6OWbtt27ZpxIgRaty4sTw9PeXq6qqqVavq2Wef1ZEjR/Ltf/HixQoMDFSRIkVUpkwZhYaGau/evbY+LAAAAAAAgGtGEuoGS09PV/v27fXDDz+oWbNm6tChgxITE/Xyyy9rwIABZm0/+OADTZw4UZLUunVrPfjggzIajfriiy/UpEkTq4moL7/8Ut26ddP69evVtGlTdejQQZs3b1azZs20f//+m3KMAAAAAAAAheVk7wDuNLGxsWrQoIH27dunsmXLSpL279+vtm3basaMGQoJCVFISIgkadCgQZo8ebK8vb1N2+fk5GjcuHF66623NHLkSH333XemusTERL300ktydnbWkiVLdP/990uSMjMz1b9/f82cOfPmHSgAAAAAAEAhMBLKBsaPH29KQElStWrVNGrUKEnS1KlTTeXt2rUzS0BJkoODg0aPHq0KFSpo8eLFZnXfffed0tLS1KtXL1MCSpKcnZ01efJkFSlSpFBx+vv7W/1hRBUAAAAAALjRGAl1g5UuXVodOnSwKO/Vq5eGDBmimJgY5eTkyMHhcv4vOTlZixcv1q5du3TmzBllZ2dLujy6KTk5WadPn1bp0qUlSVFRUZKkxx57zKL/MmXKqGPHjlq0aJGNjgwAAAAAAODakYS6wXx9fa2WlyhRQiVLltSZM2eUkpKiMmXKaM6cOXrmmWd04cKFfPs7f/68KQmVu0ZUfvvw8/MrVKxxcXFWy/39/QvVDwAAAAAAwNUwHc9OEhMT1a9fP2VkZGjSpEnat2+fLl26JKPRKKPRqMDAQEmS0Wi0c6QAAAAAAADXjyTUDXbw4EGr5efOndOZM2fk7u6ukiVLatmyZcrIyNCwYcP0wgsvqHr16nJ3dze1P3DggEUfPj4+ki4nsKzJrxwAAAAAAMDeSELdYMnJyQoPD7conzt3riQpMDBQjo6OSklJkSRVrFjRom1kZKSOHz9uUd6mTRtJ0s8//2xRd/r0aa1cufK6YgcAAAAAALAVklA2MHz4cCUnJ5t+j4+P19ixYyVJQ4cOlSTVrFlTkjRz5kxdvHjR1DYpKUmDBw+22m///v3l6uqqWbNmafXq1abyzMxMvfTSS2b9AAAAAAAA3EpIQt1gLVq0kIODg6pXr67Q0FB17dpV9erVU1JSkvr27asePXpIkrp27Sp/f39t2rRJ1atX18MPP6zOnTurZs2aKlWqlFq2bGnRd5UqVfTJJ58oMzNT999/v9q1a6devXqpZs2a+uWXX9SnT5+bfbgAAAAAAAAFQhLqBnN1ddWaNWvUu3dvxcbGasWKFapUqZLGjx+v6dOnm9q5uLgoKipKQ4YMkZubm3799Vft3r1bzz//vFatWiVnZ2er/Q8dOlRhYWFq2rSp1q9frxUrVuiee+5RbGysqlevfpOOEgAAAAAAoHAMRh6/hiv4+/tLkuLi4uwcyQ1Uw2DvCAAAuG2lOzprWNdRkqQp416Tq4uLnSMCAAC3I0ZCAQAAAAAAwOZIQgEAAAAAAMDmSEIBAAAAAADA5khCAQAAAAAAwOZIQgEAAAAAAMDmnOwdAHBT7OMhkAAAXLOMDGnkh/aOAgAA3OYYCQUAAAAAAACbIwkFAAAAAAAAmyMJBQAAAAAAAJsjCQUAAAAAAACbIwkFAAAAAAAAmyMJBQAAAAAAAJtzsncAwA1Xw2DvCAAAuLM4OktdR9k7CgAAcJtjJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJNRtLjg4WAaDQQkJCfYOBQAAAAAAIF8koQAAAAAAAGBzJKEAAAAAAABgc3dtEiotLU1ubm7y8/OzqAsJCZHBYFDr1q0t6po0aSIHBwedPHnSVPbnn3+qT58+8vHxkYuLiypUqKAnnnhCe/bssdg+IiJCBoNB/fr107FjxzRw4EBVrFhRTk5OmjRpkiQpIyNDn3/+uZo2baoyZcqoSJEi8vPzU+fOnTV37lxJUkJCggwGg/744w9JUpUqVWQwGEw/AAAAAAAAtxInewdgL25ubmrevLkiIyOVkJBgSkbl5OQoMjJSkrRx40ZdunRJRYoUkSSdPXtWW7duVd26deXp6SlJCg8PV5cuXZSamqpGjRopODhYf/31l3788UeFhYVp2bJlatOmjcX+T548qaZNmyorK0utW7dWWlqaaT99+vTR/PnzVaxYMbVp00bFixdXUlKS1q5dqwsXLuixxx6Th4eHnnzySS1fvlzHjx9XaGioPDw8bsKZAwAAAAAAKLy7NgklXV7UOzIyUhEREerXr58kafv27UpJSZG/v7/i4uIUExOj++67T5IUGRmpnJwcBQcHS5IuXryoPn36KDU1VVOnTtXQoUNNfU+cOFEvv/yyevfurX379snNzc1s38uWLVP37t01e/Zss7r4+HjNnz9fvr6+2rx5s8qUKWOqS0tL09atWyVJZcuW1fTp0xUcHKzjx49r/PjxVkd1AQAAAAAA3Aru2ul4kkzJpIiICFNZ7r9Hjx6db11QUJAk6eeff9bx48cVGBholoCSpJdeekmNGzfW4cOHtWDBAot9u7q66tNPP7VITuVO82vUqJFZAkq6PHorMDCwUMf4X/z9/a3+7N+//4btAwAAAAAAQLrLk1AtWrSQq6urRaKpWLFiCg0Nla+vr9UkVG7yKioqStLl6XPW9O3b16xdXgEBAapQoYJFee3atVW0aFEtXbpUH3/8sY4cOXINRwYAAAAAAHBruauTUO7u7mrWrJkSExOVkJCgnJwcRUVFqU2bNnJ0dFRwcLBpXaizZ89q27ZtZutB5SaI8psGl1uelJRkUVe5cmWr2xQvXlxff/21XF1dNWLECFWoUEG1atXS4MGDFR0dff0HnUdcXJzVn2rVqt3Q/QAAAAAAANzVSSjJfEpe7npQuWXBwcHKyMhQTEyMaT2o3Kl4BfFfT6m7chpeXr169dKBAwf09ddf65FHHtGZM2c0bdo0tW7dWq+88kqB9w8AAAAAAHCruOuTULlJpYiICIvpdnkTVFfWSVL58uUlSYmJiVb7TkhIkCSr0+6uxtPTUwMHDtTPP/+sY8eO6bffflPx4sU1YcIExcXFFbo/AAAAAAAAe7rrk1AtW7aUi4uLKdFUvHhxBQQESLo8nS53XagrFyWXpDZt2kiS5syZY7XvmTNnmrW7VgaDQZ06ddJDDz0kSWZJKBcXF0lSVlbWde0DAAAAAADAlu76JFTedaFWrlxpWg8qV3BwsDZs2KBt27apdu3a8vb2NtX17NlT3t7eWrt2rb766iuzfqdMmaJNmzapQoUKCg0NLXA8W7du1cKFC5WRkWFWfvr0aa1fv16SVKlSJVN57misPXv2FPygAQAAAAAAbjInewdwKwgKCtLatWuVlpZmNt1OupyEmjFjhunfeRUtWlSzZs1Sly5dNGjQIH311VeqWbOm/vrrL23dulUeHh6aM2fOf67/dKXExESFhoaqRIkSatKkicqVK6czZ84oMjJS58+fV5cuXRQYGGhq37VrV82YMUO9e/dWx44dVaJECUnSN998c20nAwAAAAAAwAbu+pFQknlyyVoSKr86Sbr33nu1ceNG9erVS4cPH9b8+fN17Ngx9e3bV5s2bSr0VLwWLVpo3Lhxaty4sfbs2aN58+Zp06ZNatCggb777jstWLDArH2PHj00ceJEVaxYUUuWLNG3336rb7/9tlD7BAAAAAAAsDWD0Wg02jsI3Fr8/f0l6fZdAL1G/k8lBAAAhZfu6KxhXUdJkqaMe02u/6xJCQAAUBiMhAIAAAAAAIDNkYQCAAAAAACAzZGEAgAAAAAAgM2RhAIAAAAAAIDNkYQCAAAAAACAzTnZOwDghtvHAx8BALihMjKkkR/aOwoAAHCbYyQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbM5gNBqN9g4Ct5ZixYopMzNT1apVs3coAADgFmA0GnXsxClJUjmvsjIYDHaOCAAA2FO1atW0ePHiQm/HSChYKFq0qJydne0dxl1j//792r9/v73DAO4oXFfAjWUwGHTpwjldunCOBBRwg/GZBdx4XFe3LkZCAXbm7+8vSYqLi7NzJMCdg+sKuPG4rgDb4NoCbjyuq1sXI6EAAAAAAABgcyShAAAAAAAAYHMkoQAAAAAAAGBzJKEAAAAAAABgcyShAAAAAAAAYHM8HQ8AAAAAAAA2x0goAAAAAAAA2BxJKAAAAAAAANgcSSgAAAAAAADYHEkoAAAAAAAA2BxJKAAAAAAAANgcSSgAAAAAAADYHEkoAAAAAAAA2BxJKMAOLl68qB9//FHPP/+8mjdvLldXVxkMBo0ZM+aq2x4+fFj9+/dX+fLl5ebmppo1a+qtt95SWlqa7QMHblMREREyGAz5/rRo0cLeIQK3rNTUVI0ePVo1a9aUm5ubypcvrwEDBigpKcneoQG3reDg4P/8XFq+fLm9QwRuSZs3b9YHH3ygHj16qGLFiqZr5mqmT5+uZs2aycPDQ6VLl9aDDz6omJiYmxAxruRk7wCAu9G+ffv0xBNPFHq7v//+W4GBgTp16pTq1aunNm3aaNOmTRo7dqzCw8MVHh4uV1dXG0QM3BmqVaum1q1bWy0HYCktLU3t27dXbGysfHx81K1bNyUkJOj777/Xr7/+qtjYWFWtWtXeYQK3rdDQUHl4eFiUV6hQwQ7RALe+d955R7/88kuhtnnxxRc1efJkubu7q2PHjkpLS9OqVau0cuVKzZ8/XyEhIbYJFlaRhALsoFixYnrqqafUtGlTNW3aVEuXLtXo0aOvul2/fv106tQpDRs2TJMnT5YkZWVlqWfPngoLC9P7779foNFUwN2qdevWmj59ur3DAG4b48aNU2xsrAIDA7Vy5UrTl+UJEybolVde0YABAxQREWHfIIHb2Pjx4+Xn52fvMIDbRmBgoBo0aGD6HuXn56f09PR8269evVqTJ09WmTJltG7dOtWoUUOStG7dOgUHB6t///4KDg5WyZIlb9IRgOl4gB1Uq1ZN33zzjQYNGqSAgAA5OztfdZsNGzYoOjpaXl5e+uijj0zlTk5O+uKLL+Ts7KwpU6YoKyvLlqEDAO4SGRkZmjp1qiTps88+Mxut8fLLL6tBgwb6448/tHnzZnuFCAC4y7z22msaO3asunTponLlyl21/YQJEyRJI0eONCWgpMvJrMGDB+vMmTP69ttvbRYvLJGEAm4TS5culSR16dLFYsqdt7e32rRpo5SUFK1du9Ye4QEA7jDR0dE6e/asqlWrpkaNGlnUP/zww5KkJUuW3OzQAAC4qtTUVK1Zs0bSv59ZefE5Zh9MxwNuE9u3b5ckBQQEWK0PCAjQmjVrtGPHDgUHB9/EyIDbx759+/S///1PycnJKlu2rFq3bq1OnTrJwYG/yQBXKsjnjiTt2LHjpsUE3Gm+/fZbJScny8HBQTVr1lRISIgqV65s77CAO8KePXuUnp4uT09PVaxY0aKezzH7IAkF3CYOHjwoSVZvoHnLExMTb1pMwO0mJibG4kko9evX14IFC8yGaAPgcwe4GcaNG2f2+/DhwzVq1CiNGjXKThEBd46rfY4VLVpUJUuWVEpKis6fP69ixYrdzPDuWvzpF7hNXLhwQZJUpEgRq/VFixaVJJ0/f/6mxQTcLkqUKKFXX31VsbGxSk5OVnJyssLDw9WiRQvt3LlTHTt21NmzZ+0dJnBL4XMHsJ22bdvqxx9/1P79+3Xp0iXt2bNH7777rpycnDR69GjTA2gAXLurfY5JfJbZAyOhgGvQvXt37d69u1Db/PDDD2rWrJmNIgLubNd7zTVq1MhiTZv27dtr7dq1ateunaKiovT555/rf//73w2LGQCA/IwdO9bs95o1a+qNN95QkyZNdP/992vMmDF65pln5O7ubqcIAcA2SEIB1yA+Pl579uwp1DaXLl26rn3mPpUov34uXrwoSQwjxR3JVteco6OjXnvtNUVFRWnFihUkoYA8+NwBbr6OHTuqSZMm2rRpk9avX886n8B1uNrnmMRnmT2QhAKuwbZt2276PitXrqytW7fq8OHDVutzy319fW9mWMBNYctrLnctqKNHj9psH8DtKHdxZD53gJurRo0a2rRpE59LwHW62ufYxYsXdebMGZUqVYok1E3EmlDAbeKee+6RJG3ZssVqfW55gwYNblpMwJ0gJSVF0r9rAgC4jM8dwD74XAJujFq1asnV1VUnT55UUlKSRT2fY/ZBEgq4TTz00EOSpCVLlig9Pd2s7vjx44qKilKpUqXUqlUre4QH3LYWLFggKf/H0AN3q1atWqlEiRLav3+/1dGI8+fPlyR16dLlJkcG3LlOnjypqKgoSXwuAdfL3d1d7du3lyTNmzfPop7PMfsgCQXcJpo1a6ZWrVrpxIkTeu2110zlWVlZevbZZ5WZmalhw4bJ2dnZjlECt6ZJkybp0KFDZmVGo1HTpk3TxIkTZTAYNGTIEDtFB9yaXFxc9Nxzz0mShg4dalo3Q5ImTJigHTt2KCgoSI0bN7ZXiMBtKSYmRosWLVJ2drZZeUJCgrp3766LFy+qa9eu+T5WHkDBvfzyy5KkcePGad++fabydevWadq0aSpZsqSeeuope4V3VzIYjUajvYMA7kbdu3c3zfU/cuSIDh06pAoVKpj+w+Hj46OwsDCzbfbt26fAwEAlJyerfv36qlu3rjZu3KgDBw6oZcuWWrNmjVxdXW/6sQC3Oj8/Px0+fFgBAQGqUqWK0tLStHPnTsXHx8vBwUGTJ082fdkG8K+0tDQFBwdr/fr18vHxUZs2bZSYmKj169fL09NTsbGxqlq1qr3DBG4r06dPV//+/VWuXDkFBASoZMmSSkxM1ObNm5WWliZ/f3+tWbNGXl5e9g4VuOUsXbpU77zzjun3DRs2yGg0qnnz5qayUaNGmWaRSNKLL76oyZMnq0iRIurQoYMyMjK0atUqGY1GzZ8/XyEhITfzEO56JKEAO/Hz81NiYmK+9b6+vkpISLAoP3TokEaPHq3ly5fr9OnTqly5snr16qU33nhDbm5uNowYuH19+umnWrlypeLi4nTixAllZmaavlAPGzZMTZs2tXeIwC0rNTVV77//vmbPnq1Dhw6pdOnS6tSpk9555x1GagDXYPfu3fr000+1fv16HTp0SCkpKSpatKjq1KmjRx55REOGDJG7u7u9wwRuSblJ3P/y/fffq1+/fhbbTZ06Vbt375aLi4tatGihUaNGqWXLljaMFtaQhAIAAAAAAIDNsSYUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQAAAAAAABsjiQUAAAAAAAAbI4kFAAAAAAAAGyOJBQA3MW8G4fIu3GIvcMw0/2ZN+XdOEQHjxy32T6adH76ljvum+GP2G3qMuB/qtrmMYvX/lTKWT03epLqd+wvn6Y95N04RHMXh0u6/D5p0vnp69r3sLcmy7txiKI37byufmBfd+u1czNcy71v7uJweTcO0cfT5tgwMtwqrvU+yv0XwK3Eyd4BAABwIx08clxNuwxSy8b+CvvqXXuHc8s4fPSk+g3/QJmZWWrbvIHKliphVv/S2KlaGblRdWv4qU2zBnJydFCVSj52ivbGa9L5aR06elLHNy+ydyi3LK4d4PbE/Q3A7YQkFADglvLp2y8qNS1dPp5lbLaP+V+MVWZWts36vxVFbtiuS6lpenlgT702pLdZXUZmpsKjN6tSeS+Fz54gBwfzgdJr50+Vs5Pjde3/zece1/P9QlWhnOd19QP7uhuvnZvlZtz7cHu71vso918AtxKSUACAW0pFH9v/J9nvDhrhU1BHjydLknwreFvUnTh1RtnZOark42mRgJKkGlUqXvf+vT1Ly9uz9HX3A/u6G6+dm+Vm3Ptwe7vW+yj3XwC3EoPRaDTaOwgAgH3kru1S0CH8ScdOauK38/R7zFadSE5RMY+iat6wjob1D1Uj/xpWt1m6Zp2mzgjT7n0JcndzVasm9fXm849r/rIIjf/qJ01+63k91vVeU/vuz7ypmM1x2rhkmiqX/zdhcujoCX36/UJFbdyho8dPydXVRd5lS6lZwzoa3KebqvtV0MfT5mj8Vz9ZjePRzu005e0XJP331IWkYyf12Q+L9Pu6rTpy/JTcXF3kW8FbHds21aDeXVXMo0iBztXpM+f05azFWvHHBiUmHZejo4MqlvPUva0CNKh3V7MvBFlZ2Zo+/zf9tGSN/k48IkmqWbWSenVpr8d7dJSjo+UopKysbM0MW6mfl0Zoz4GDysrKVjXf8nq0S3s91fMhOf0zcil60071GDTKaozDn3lUPy1Zo0NHT1rUVfLx1KZfv5Z0+X2S9/e89sYf0uc/LtLaDTt0/NTl90TVyj56MLiFnundxRTHsLcm66dff9fCae+oVZP6Zn1cSk3X13OWaPHqaB04eFSSVLtaZfV7uJMe7dLeYp+58az/5Ut9/uMvmv3Lah0+ekJlS5dQj05tNWJwb7m6OF/1+PM7poLKO31t9pTRGv/VT1q0IkonklNU3rus+nbvqOee7C6DwWCxbcrZ8/rsh0Va8cd6HTxyQs5OTmpQp5oG9+mqjm2bWt1fYa+lXXsOaOHySEVt2KGkY6d0/uIllfMqo/aBjfTSwJ4ql+c9eK3Xzo7d+9Wh7ytq5F9Dy3/42Or238z9VW9+/I0G9emqsS8PMJUX9D1cGPviD8u7bCkVL1a00NteTdKxk5o6I0zh0Zt17ORpubu5qkmD2npxwMNqek9tU7ula9ZpwKsfqmuHVvr6g1et9vXWhO/05azFGjd8oJ7u1VlS/vc+Sdqwbbc+mjZHW3btlaODgwLq1dSIwb21L/6QXnj7Uw1/5lG9OqiX2TaFPb95X9uZYav0zdxfFX/oqIp5FFGnoGYaNewJlSjmYXEsmZlZmv3Las1f9of+2p+ojMwslfMsrRaN6mrAow/qnjrVzdrvjT+kqdMv38tPnT6r4sWKqlWT+np5YE/Vrla54C9IAe2NPyQfzzIFvnfnynvPSkvP0KRv52vX3ng5OzmqVZP6emNo33wT9POW/q4fFqzUn38nKDsrW36VfNT9/jYa1Ker3FxdzNpmZGZqVtgqzV2yRolJx5WWlq6ypUuoVjVfPfxgkLrf38ZqTK2a1C/w/e3K7TIzs9Tg/v66mJqmnSu/t/q6/rkvQe0ee1H1a1XV6tkTzOrWxGzRt3OXamvcPtN95YHg5npxwMMqXbJ4oc4zgLsPI6EAAAXy574EPTx4tJLPnFN13wp6sH2gko6d1LLfY7UycqO+ePdlde3Qymybr2Yv0ahPvpWDg4NaNKorr7KltGXXXnV64lV1bGP9i7Y1ScdOqkOfV5Ry9ryqVi6ve1s3VnZ2jg4fPamZYavUpH4tVferIP+aVdT53kD9Gr5OnmVKqn1gI1MfzRrWvep+YrfG6YmX3tPZ8xdVqbyXOrZpotT0DP2dkKSPp81Vp6Bmqler6lX72Rt/SI8OHaMjx5PlVaaU2v0Tx/6DR/TZD4vUpEFtPdiuhSQpOztbT778nlZHb1axokUU1PweGY1Grd20U699ME1/rN+ubz8aYTZCKTUtXX1eGKfoTTtVqkQxNa5XU66uLtqya69Gf/Kdojft0vTxr8vBwUFeZUrp0c7ttGtvvOL2JqjZPXVUpVI5SfrnfLXUoaMnLM5ZQb5ILF4VredGT1J6RqZqVqmoB9q10PkLl7Rn/0G9PXm6+nS/z+qXm7xOnj6jns+O0Z/7EuRVppRaBvjLKKM2bv9Lw8ZM0bbdf+v9Ec9Y3XbwmxMUHr1ZLRvXUzXf8lq/9U9NnRGmoydO6/NxL0mS6fiXhK/TpdQ0Pdq5nWn7vMeYm1CSCp6UzZWRmaVHh47R3vjDatnYX5dS07VuS5zGffqDLlxK1f+e7WPWfn9ikh4Z8paSjp9SpfJeCm7RUBcupWrzzr16/KV39dYL/fTsEyFm21zLtTRl+kItDV+nujV81axhHUnSrr3xmj5/uX6LWK+VMz8xJaKu9dppUKeaavhV1Na4fUo4dNTqSKkFv0VKkh5+IMhUVpj3cEHtOXBI3Z8ZqSqVymnxN+9ZTd5eq407/lLfF8bpzLkLqu5bQfe1bqzklHOKWLdVv6/bos/HvayQjq0lSfe1bqLiHkW0KnKjLl5KVdEi7mZ95eTkaNHKtXJ0dFDI/a2vuu+VkRvVf/gHysrOViP/GvKtWE5/7o1XyNNvWE3SStd3fsdOnq6v5/yqlo3rqUolH23Yvls/LlypffGHtejrd82SqhdT09Rn2DtatyVORdzd1LxhHRUvVlSHjpzQgt8iVcyjiFkSatnvsRr8xidKz8hUvVpV1LheLSUdP6XFq6K1MnKj5nw6WoEB/gV6TQri0NETCnn68nvip6lj5FHU/eobXWHJ6hhNn79c99Stro5tm2r3vgQt+z1Wazfu0KKv35V/zSpm7Ye/+7l+XLhSbq4uat2kvtzdXBWzeZfe+2ymVkZt1LzPx6qIu6up/bNvTtSS8Bh5FHVXi4Z15eFRRMdOJGvDtj91KTXVLAl1pYLe367k7OykLve10owFy/Vr+Dr1Celg0cZ03T4YZFb+zpQfNHXGQrk4O6lh3RryLltKcfsSNO2fP7os+e4DeZUp+Z/nFMBdzggAuGt5BXQzegV0u2q7nJwcY1DPYUavgG7GsZOnG3Nyckx1S1ZHG8s16W6s0vpR47ETyaby+ENHjRWbhxorNg81Rm3YYSrPzMwyDhszxbTvOb+sNttXyNNvGL0CuhkTk46Zyj78YrbRK6Cb8fUPp1nEdujICWP8wSOm3xOTjhm9AroZQ55+I9/jafzQQIvjPn3mnLHOvU8YvQK6GT+bEWbMzs42q9+wfbfxRHJKvn3mPb6W3Z81egV0M44c/40xPSPDrH7334lm8X7+Q5jRK6Cbsc0jzxuPn/q3/2Mnkk39fDP3V7M+Xnv/S6NXQDfj069/bDx77oKp/PyFS8bez481egV0M06f95vZNh99Odvq+TYar37OvAK6GRs/NNCsbH9ikrFy4CPG8k17GOcvizCry8nJMf4es8WYlv7vsT8/epLRK6Cbce3GHWZtc+N98+OvzdofP5Vi7Nj3FaNXQDdjePRmi3i8AroZW4cONR4/edpUnnD4mLFGUG+jV0A3s3NsNFp/za2dg4JcD9a2CXn6DeO58xdNdVvj9hl9mnY3+rbsabxw8ZKpPCsry3QtTZ2x0Ox9duDgEWOTLs8YfZp2N/65L8FUfq3XUtSGHWbvKaPRaMzOzjaOnzbX6BXQzThszBSrx1PYa+eTr38yegV0M37y1VyL9vEHjxi9AroZW/UYalZ+Le/h/7L3wCGjf4cnjd6NQ4yzFq0q8HYFce78RWP9jv2MPk27G+ctNX+vb43bZ6wZ3MdYpfWjxpOnz5jKX3z7U6NXQDfjz7+usejvj/XbjF4B3YyPPfe2Wbm1e9/5C5dM96XZeV7fnJwc49jJM0yv/Udfzjbr61rOb+5rW69DP+O++MOm8lOnzxqbdxts9AroZoxcv91sm9zj7DbwDbPjNxovX8Obduwx/Z6YdMzo1+pRY5XWjxojYreZtQ2P3mys0CzU2OiBpyzumdcjOzvbdO/pMuB1s2vxanK38wroZvxhwQpTed5z377Xi2bbLFkdbfQK6GZscH9/4/7EJFP52XMXjF0GvG70CuhmfGvCd6byhMPHTPfX5JSzZn2lpqUbN2zfbTWmK++jV7u/Wdsudkuc0Sugm7HHoJEW7XNycowBDw40lmvS3Xg0z+f6LyvXGr0CuhnbPvK88UCee2xOTo7pc/rp1z/ONw4AMBqNxoL/eQkAcNeK3rRLu/9OVMVynnp9SB+zv4R3vrelHghupouX0jR7cbipfM7icGVkZunhB4PVuum/06+cnBw19uX+KlrErcD7T045K0lq2+wei7qKPp43ZJ2aWYtWKTnlrNq3DNCzT4RYjBBo2qC2PEuXvGo/S9es09+JSapVrbLGvNhPLs7OZvW1q1U2i/ebuUslSWNf6m/212Nvz9Ia/eKTkqSv5/xqKj95+oxmhq1SBe+ymvzW82bTjjyKumvi6Ofk4uyk6fOXF/jYr8W0WYuVlp6hPiEdFPqA+V/KDQaDggMbmabE5WfXngNaHb1ZDf1raOzLA8zae5UpqfFvPitJ+R7LuFefllfZUqbffSt4K/Sfv9rHbv2zUMfj7OSk6r4VVN23QqG2kyQHBwd9/MazZtN9GtatrvYtA5Salq5tf/5tKl8RuVG7/05U53sDNfSJ7mbvsyqVfPT2i/2VnZ2jmWGrTOXXei21blrfYkSCg4ODXnnmUfl4ldHKPzYU+lityX39Fy6PsqjLHU3R44G2prIb/R7en5ik0MGjdOr0WY1/81n17nbf9RyOhdmLV+v4qRQ906uLxaiQhnWr66WBPXXxUprmL/vDVJ7bLvf488otC81zTvKzJDxGySlnFRjgr155ploaDAa9PqS3yntbLmJ+vef3tSG9Vd3v3+ugTKniejL0fkmXR4vmOnbytH76dY1cXZz16dgXLJ646VWmpBrXr2n6/avZS3QpNU1vPve4gpqb38vbtwzQkw93UtLxU1odtfmq56WgHBwcNOmt59XzoXZav223+rwwTpdS0wvVR9N7auvxHh1Nv+c997v2xGt9nntN7v38lWceVdXK5U3lxYsV1fuvD5LBYNAPC1coLT1D0r+fbfVqVbUYueTm6qKmDWrLVpo1rKNK5b0UszlOx06eNquL3fqnDh87qVZN6plN25303TxJ0pfvvWL29FSDwaBXBz2merWq6NfwGCWnnLNZ3ABuf0zHAwBcVe5/srt2aCVnZ8uPjocfDNbSNbFm/xnfsG335W3ua2XRvkQxDwW3aKila2ILtP8GdapJkt7/bKYcHRzUtvk9FutqXK/I9dslyezLxjX1s+FyP31C7rvqdKDDR0/q8LGTKlOqhILzTH/K1bFNU5UoVlTxh47qxKkUeZUtpZhNu5SZlaV2LRvJ3c3VYhuvsqVUpXJ57f47Ualp6Vbb3AiRG3ZIkp4IvfbzFRG7TZL0QHAzq9OC6teuqqJF3LR11z6LOmcnJ7VuUs+ivNo/X/yOn0opVCw+XmUUvfCzQm2Tq6KPp9mX9ryxrLoilj/+Oebc6ZhXat7o8tS3rXH/HvP1XEunz5zTisiN+uvvRJ29cFE52TmSpMysLJ0+e14pZ8+rVIliVz/I/+BbwVtN76mtjdv/0o7d+03XqyQtWH45MZM3UXkj38Pnzl9U6ODROn4qRa4uzvrix0X64sdFBY599qejrS7Wn1fua/ZQ+0Cr9S2svGaBAf4q711GURt26OTpM6YEdlp6hpaGr1MRdzc9kM97IK/cZGruVL+8nJ2d1Pnelvpq9hKz8us9v8EtGlpsU9XX8rqK3rRT2dk56tC6iSr5eF31WApyHr+Z+6u2xO3Tg+3/+9wsWrlWH38556r7zJXzzxK467bE6Zn/fayZk0YWeNurnfvYrX+qeaO6yszM0uadeyTJIjEvSf41/FS3hq/i9iZo1954NalfSzX8KqqIu5tWr92sz34IU+gDQWZJH1syGAzq0amtJn83X4tWRGlw326mugW/Xb5u806hPXn6jOL2Jqhq5fKqU93Xan/N7qmjXXvitWP3frVrafmZBgASSSgAQAEcO3X5r6SVylv/olH5n/KjJ5JNZSf++bJSvlxZq9sU5lHRj3Vpr4jYbVq8KlqPv/Su3Fxd1LBudbVr2Ui9u95nNhrmWh355+lxfhXL3bR+TOc1n6diGQwGVfTx1NnzF3X0ZLK8ypbSoaMnJEkzw1aZjZax5sy5CzZLQh05fkqS5Ffx2kehHTpy+Vje/2yW3v9sVr7t0tMzLcq8ypS0muTz+Gf9nYxMy21spbyX5WgUSab1ZzIy/o0l9/V7duREPTtyYr59nj7z70iCa72WFi6P1PB3P9fFS2n57ufCpdTrTkJJUminttq4/S8t+O0PUxJq259/a3/iETW9p7ZZoudGvoczs7J17sJFSVJ6Rqb+TkwqVNx5X5v85L5POw94/T/b5X3NHBwc1P3+Nvrsh0X6ZeVaDXzs8uLjq6I26fzFS+rRqa2Kul99NOjxf0ao5PfkPGvJn+s9v9ZGV5muqzznq7D3zNzzeE+nAf/ZLu95zM/ZcxcK/Vr/2//5QrWvmE+CLffc597HU86eV0ZmlsqULJ7va1vJx0txexN07J/PymIeRfTJyGc1/N3PNXbyDI2dPEPVfMurVZP6euTBYNNabrYS+kCQJn83Xwt++8OUhMrIzNSS1TFyc3UxSxjmvn4HDh4xPdQkP8kFeA0B3L1IQgEArpu1p3/dSI6Ojvr6g1c1rF+olv+xXms37tSWXXsVu/VPfTp9oeZ++pbZ06nuJFee25ycy3/Rr1erivxr+P3ntldOBbzV5I5OaN6wTqGTfw4Otn3PFUZhYsl9/dq3DJBn6RL5trveJ0wdOnpCL4yZIkl655Wn1KFNE5XzLG1KODzU/zVt2rFHxhv0kORuHVtr1CffadHKtXrrxX5ycHAwjaYI7WQ+7exGvofLlCquWZNHqdfzYyWjUT9MetPqtN3rkRtvl3tbmi0ofaXqfuZPSgt9MFif/bBIC5dHmZJQpnNSgKl41+p6z29hFoMvjNzrPe/i2dYE1Kv5n/WS9OTDnfTkw50KvO+J387TB5/PUjXf8vr+49cKvN2NZu2zskentmrb/B4tj9igP2K3KmZLnH5YsEI/LFihwX266u2X/ztpdz1qVa2k+rWqasdfB/R3QpKq+1VQePQWnTl3QV3va2k2xTj39bv8sI2G/9lvfn9YAQCJJBQAoADKlb08PeDwP39hv9LBf/5C6pNnRIhX2VL6OzFJR46dUq2qlSy2yR1JUxj1a1dV/dpV9eqgXjp/4ZI+/mqups1arJGffKsV+TwevqDKe5fRvoTDSjh8THWv8sXtav1IUsLhY1dtm3teDx09mW+bw//U+XiWMeu/WcM6+T417mYo711WBw4eUcLhowV6YqA1ue+XB4Kba8jjITcwultX7jH3CblPne9tWaBtruVaCl+7WRmZWXr28RA907uLRX1i0vFCRv7fSpcsruDAhloVtUnRm3apZWN/LVqxVs5OTup2xXSmG/0eDgzw18xJI9X3hXf0xIvv6sdJI9WmWYPr7jeXj3cZ/Z2YpOf79zB70tvV+NfwU+1qlbV55x4lHD6mksU9FB69+fL02xYFm6rk/c8oz8P53CMOH7Msv1n3iMLc66TL7/2Ew8c05qX+151kLYzJ383XB5/PUpVKPlrw5TvyLuR0t/w+93LPfe59vFSJYnJxdlLymXO6mJpmdTRU7ii1cleMnixbqoT6du+gvt07yGg06vd1W/XM6+P15azF6tXtPtWuVrlQMRdG6ANttXPPAS347Q+9NqS3FuYmSh8MNmuXO+KzdMlimvL2CzaLB8Cdj4XJAQBXlbtOzeLVMcrOzraoX/DPgry57SSZphH8uibGov258xdN6wFdq2IeRfTmc4/LYDBoz9+JpvLcv+xnZeUUqr+2/yyUOzNs5XXFlTsKY/ai1crJ+e8YKvp4qmI5TyWnnDWtJZXXqqhNOnPugqpU8jFNOWzVpL4cHR20KmqTMjOzrivW69H2ny/5Py689vMV1LyhJGnZ7+tvREj/ydn0vrB8/95MQf+st1OYY76Wa+nM+QuSzBPDudZtidPJ5DMW5dd67eT6d4HySK3duFMnklMUHNjQIuFgi/dw66b1NWPimzJKevPjr63ep67V9bxP856TJatjlJGZpW4dW8nJ6b/Xi8uVe0/9ZVW0RV1WVraWhq+zKL9Z94jc/UTEblOSlWTYlW7m9Z4rMem4Jnzzs3wreGvBl2OtXg9Xs/gq5755o8vXp7OzkxrXryVJWrTCcpH+3X8nKm5vgooWcVO9mlXy3Z/BYFD7lgG6r3VjSdKe/QevGuP13N+6d2orBwcHha2I0vkLl7QyapNKlSime1sGmLUr711WNfwqam/8Ye2/xqmQACCRhAIAFECrJvVUp7qvDh05oQ+/nGM2hWfZmlgt/T1WRYu4qXeepzf16tJeLs5Omrc0Quu2/PtEpezsbL018XtduJha4P3PW/q7dudJNOUKj94so9FotlZO6ZLF5OzkpITDxwr1RbRPSAeVKVlc4dFbNG32YotpSpt27tHJ02eu2s9D7QNVzbe8/tp/UGMnz7D4EvjX/oNmIweeevQhSdJbE77XqX+elCRdXgdo7OTpkqSne3U2lft4lVGvrvfq0JETGvzmJzphJZkQf+iofg23TFjcSM/07iI3VxfNDFulRSvXmtUZjUb9EbtN6VdZb6dx/ZoKan6PNmzfrdc/mKbzFy5ZtInbG681MVuuO95ynpeTePmtI3P0RLJa9RiqVj2GXve+/kvn9oGqVbWSFvz2hyZ8/ZPFOTIajdqwbbdpMXLp2q6lapUvL5S+4Lc/dDH13zWhjp5I1oj3vrAa27VeO7k6BTWTR1F3LV2zzvSkTGsLNNvqPRzU/B79OOlNzZw88qoPBSiMJ0I7qmzpEvpsRph+WLjCIrmclZWt32O2Wr1H9ejUVgaDQWHLI7XQyiLtV9P1vlYqXaKYYjbv0k9L1pjKjUajPpo2x+pIqJt1jyjnWVqPPNROaekZev6tKRZrOZ08fUabd+41/T7k8W5yd3XR25O+19I1lsmz9IzLaxFdyyjZ/PhW8NYPE9/UgmnvFGodwrzWb9ut2b+sNivLPfd1a/ipRSN/U3nu/Xz8V3PN7vMXLqbqjY++ktFo1BM97jc9WGPnXwe0dM06izXsUs6e15Zdl89dfmvB5XW1+9t/b1tarZrUU/yho3pnyg9KS89Ql/taWn0IyUsDH1FOTo4GvPqhdu05YFF/+sy56/rDBIC7A9PxAAB64MkR+db1Cbk8ReDzcS8pdNAoTf5uvn77PVb+Naso6dgpbdi+W06Ojpo46jmzaQ5+lXw0atiTGvXJt+oxaJQCA/zlWaaktu7aqzPnLujhB4M0f9kfVv+je6Vfw9fpudGT5VexnOpU95Wbq4sOHjmhLbv2ysHBQa8P6WNq6+LsrHYtG2ll5Ea16/WSGtSuKmcnJzVrWMfsEedXKlWimL7+8FU98fJ7Gv3Jd/pm7lI1rFtdaekZ2hd/WPGHjip89gTTU67y4+TkqG8/ek09nx2jL2b+ooXLI9WkQS0ZjZcXdP1r/0F9P/510xpIg/p00dpNOxQevUWBIUPUumkDGY1GRW3coQsXU/VAcHP1f+QBs32MGz5Qh46c0K/h67QmZqvq1ayiCuXK6lJauvYeOKT4Q0fVKahZgad7XYtqvhU06a3n9fzoyRr0v/Ga8PVPqlPDT+cvXNRffx9U0vFT2hsxU64u/72mz2fjXlKv58bq+3m/aeHySPnXrKJynqV17sIl7d6XoKTjp/R0r85qf8Vf5Qvr/rbNFLM5Tg8PGa1WTeqriJurSpcsrlHDnpB0+Wlx17rQcWE4OTlq+if/06PPva0Pv5yjb39epro1/FS2VAmdPnNOu/bG69Tpsxr7ygDTCKhruZbuD2qqWtUqa9uff6tFtyFq1rC20tIzFb1pp+rVqmJ6ml1e13rt5HJ3c9WDwS3089LftWhFlDyKuqtTUDOrbW31Hr7R60FJl59AOGPCG3rixXf16rtfaNI381S7emWVKOahE8kp2vnXAZ09f1Hfj3/d4qlhFX081aJRXa3bEqe98YflV7GcmvwzWqYgPIq6a8Lo5/TUiA81bMwUTZ+/XL4VvPXnvgQdOHhUfbt3sLr4+M26R7w7fKD2JyYpetNONe78jAID6sqjaBEdPnpSO//arycf7qTG9S+v8VSlko++eO8VDXlzgga8+qGqVPJRjSqXnw537ESydvx1QJdS0xQ+e4LKe1898VJQQc2v7z3R7+FOevmdz/TDghXyq1hOf/6dqD37D6pY0SKaMmaYWdsu97XU4z066seFKxXcc5haNW2gIm4uit4cp+SUs2pcv5ZGDO5tan/o6AkNePVDFfcoonvqVpdXmVI6e/6iYrfG6cLFVHVs21RNG1x9vcOr3d+uJvSBIEVt2KEZC5ZLMn8q3pXt9hw4pMnfzVeHvsNVr6af/CqWk9F4eVrmn38nqKi723U/ZRbAnY0kFADA9BdXa9r/85jlujX8tGrWBE389mf9HrNVv4avU3GPInoguLmG9Q+1upjsM727qLx3GU2dEaYtO/fI1dVFrZrU05vPPa7PflgkSSpd4uprgwzu203lvctqw/bdWr/1T11KS5d32VLq1rG1hvTtpoZ1zddpmTBqqN6eNF1/xG7XwuWRys7OUXZ29lW/SLdqUl9r5kzSZzPCtGbdFi2PWK+iRdxVubyXXhvcq8BPgqtT3Vdr5k7U5z8s0orIDQqP3iIXZydVKOep557sYfYl1NHRUT9MeFPfz/tNPy1Zo9/XbZV0ecHYx7q01xOh91ssFOzu5qo5n47Wgt8i9dOvaxS3N0Fb4/apTKniqujjqUceDFbI/ZaPFb/Rut/fRjWrVtLnP4QpetMuLQ1fpxLFi6pqpfJ6pncXFXV3v2ofnqVL6tfvP9DMsJVatHKtdu05oE07/pJn6ZLyreitgY91viHHMvCxzjpz7oLCVkRpafg6ZWZlqZKPZ4G/pN1IVSuXV/isCfr252VatmadNu/co+zsHHmWKan6tarq/rZN1bWD+TEX9lpycXbW4m/e0/ufzVR4zBatitqkcp6l9dSjD+mVpx9VnxfGWo3tWq+dXKEPtNXPS3+XJD0Y3CLfJ9vdKu/hgmpSv5YifpqsabMXa9XazYrZfHlEmnfZUgoM8NeD7Vrkm+zo8UBb0wi2woyCyvVAcHMtnPaOPvpyrrbG7dO++MNq6F9dn4waqv0JSVaTUDfr/HoUddfCae9oxvwVWvDbH4rd+qdysnPk7VlaPR4IUs+HzBchfyC4uSLmTtKXsxbrj/XbFbl+u5ycHFWubGl1bNtUD7VroZpW1j2zp64dWuneVo01+fv5Wv7HBjk7OapTUDO98dzjVtdoG//ms2resI5mLFihdVt2KTs7R74Vy+mZ3p01qHdXs2uiSf1aev3ZPlq7cYf2JyZp/dY/VaK4h+pW91PvkPvyTQZd6Xrvb53bB+r1D6YpLT1DlXw8//OpfG8M7at2gY303U9LtWH7X9r990EVK+qucl5l9OTDndT1vlYF2ieAu5fBeKMeiwIAQAFlZ2er3WMvam/8Ye1Y/p1pvSMAhcO1BNjGsLcm66dff9fCae+oVZP69g4HAO4YrAkFALCZhENHdfafBZJzpWdkauzkGdpz4JDaNGvAl2agALiWAADAnYDpeAAAm1m8OkYfT5ujBnWqqbx3WV24mKq4vfE6fipFZUoWt+njw4E7CdcSAAC4E5CEAgDYTJtmDRS3L0Gbd+7Rn/sSlZ2drXKepdXv4U4a1j/0mp9WBNxtuJYAAMCdgDWhAAAAAAAAYHOsCQUAAAAAAACbIwkFAAAAAAAAmyMJBQAAAAAAAJsjCQUAAAAAAACbIwkFAAAAAAAAmyMJBQAAAAAAAJsjCQUAAAAAAACbIwkFAAAAAAAAmyMJBQAAAAAAAJsjCQUAAAAAAACbIwkFAAAAAAAAmyMJBQAAAAAAAJsjCQUAAAAAAACbIwkFAAAAAAAAm/s/6Umty+2K8dAAAAAASUVORK5CYII=",
      "text/plain": "Matplotlib figure"
     },
     "metadata": {}
    }
   ],
   "source": "# Plot interpretable term associations\nindices = np.concatenate([negative_indices[:8], positive_indices[:8]])\nindices = indices[np.argsort(coefficients[indices])]\nfig, ax = plt.subplots(figsize=(8, 5))\nax.barh(feature_names[indices], coefficients[indices],\n        color=[SCARLET if value < 0 else NAVY for value in coefficients[indices]])\nax.axvline(0, color=SLATE, linewidth=1)\nax.set(xlabel='Logistic coefficient: negative ← evidence → positive',\n       title='Strong term associations in the fitted logistic model')\nfig.tight_layout()\nplt.show()\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-083-build-a-test-error-audit-table-guide",
   "metadata": {},
   "source": "### 83. Build a test-error audit table\n\nStore row identifiers and short excerpts for inspection; export only IDs and scores, not full review text.\n"
  },
  {
   "cell_type": "code",
   "id": "step-083-build-a-test-error-audit-table",
   "metadata": {
    "portfolio_step": 83,
    "purpose": "Build a test-error audit table",
    "execution_scope": "core",
    "execution_seconds": 0.0237
   },
   "execution_count": 83,
   "outputs": [],
   "source": "# Build a test-error audit table\nerror_audit = test_df[['review_id', 'sentiment', 'text']].copy()\nerror_audit['predicted_sentiment'] = np.where(selected_prediction == 1, 'positive', 'negative')\nerror_audit['positive_score'] = selected_score\nerror_audit['wrong'] = selected_prediction != y_test\nerror_audit['excerpt'] = error_audit['text'].str.slice(0, 180) + '…'\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-084-inspect-confident-missed-negatives-guide",
   "metadata": {},
   "source": "### 84. Inspect confident missed negatives\n\nThese are negative-labelled reviews that the model calls positive. A high wrong score warns against treating confidence as certainty.\n"
  },
  {
   "cell_type": "code",
   "id": "step-084-inspect-confident-missed-negatives",
   "metadata": {
    "portfolio_step": 84,
    "purpose": "Inspect confident missed negatives",
    "execution_scope": "core",
    "execution_seconds": 0.0153
   },
   "execution_count": 84,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>excerpt</th>\n      <th>positive_score</th>\n      <th>predicted_sentiment</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>7918</th>\n      <td>this is a great movie i love the series on tv and so i loved the movie one of the best things in the movie is that helga finally admits her deepest darkest secret to arnold tha...</td>\n      <td>0.9997</td>\n      <td>positive</td>\n    </tr>\n    <tr>\n      <th>8563</th>\n      <td>this one of the best celebrity's reality shows a ever saw we can see the concerts we can see the life of britney i love the five episodes i was always being surprised by britne...</td>\n      <td>0.9883</td>\n      <td>positive</td>\n    </tr>\n    <tr>\n      <th>6491</th>\n      <td>'ernest saves christmas' is comedian ernest's christmas special film in this film ernest has to find a successor to santa claus in order for christmas to continue along the way...</td>\n      <td>0.9864</td>\n      <td>positive</td>\n    </tr>\n    <tr>\n      <th>6043</th>\n      <td>yes the movie is not a piece of art but the first time i watched it i was 10 years old my parents were out and i stayed home with my two brothers it was may 1970 i know that be...</td>\n      <td>0.9827</td>\n      <td>positive</td>\n    </tr>\n    <tr>\n      <th>9132</th>\n      <td>although this isn't a great film there's something compelling and memorable about it like another commenter on the film i saw this in childhood it's been thirty three years sin...</td>\n      <td>0.9813</td>\n      <td>positive</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                                                                                                                                                                                  excerpt  ...  predicted_sentiment\n7918  this is a great movie i love the series on tv and so i loved the movie one of the best things in the movie is that helga finally admits her deepest darkest secret to arnold tha...  ...             positive\n8563  this one of the best celebrity's reality shows a ever saw we can see the concerts we can see the life of britney i love the five episodes i was always being surprised by britne...  ...             positive\n6491  'ernest saves christmas' is comedian ernest's christmas special film in this film ernest has to find a successor to santa claus in order for christmas to continue along the way...  ...             positive\n6043  yes the movie is not a piece of art but the first time i watched it i was 10 years old my parents were out and i stayed home with my two brothers it was may 1970 i know that be...  ...             positive\n9132  although this isn't a great film there's something compelling and memorable about it like another commenter on the film i saw this in childhood it's been thirty three years sin...  ...             positive\n\n[5 rows x 3 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Inspect confident missed negatives\nmissed_examples = error_audit[(error_audit['sentiment'] == 'negative') & error_audit['wrong']]\ndisplay(missed_examples.sort_values('positive_score', ascending=False)\n        [['excerpt', 'positive_score', 'predicted_sentiment']].head(5).round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-085-inspect-confident-false-alerts-guide",
   "metadata": {},
   "source": "### 85. Inspect confident false alerts\n\nThese are positive-labelled reviews that the model calls negative. Read the full original row locally before deciding why it failed.\n"
  },
  {
   "cell_type": "code",
   "id": "step-085-inspect-confident-false-alerts",
   "metadata": {
    "portfolio_step": 85,
    "purpose": "Inspect confident false alerts",
    "execution_scope": "core",
    "execution_seconds": 0.0137
   },
   "execution_count": 85,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>excerpt</th>\n      <th>positive_score</th>\n      <th>predicted_sentiment</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>9113</th>\n      <td>honestly i was expecting to hate this one and really only checked it out because jenna jameson is in it but i have to say i got a kick out of evil breed a group of college kids...</td>\n      <td>0.0075</td>\n      <td>negative</td>\n    </tr>\n    <tr>\n      <th>8816</th>\n      <td>comanche moon had everything going for it for starters simon wincer's back a man who's name is synonymous with high quality tv westerns unfortunately the problems with moon are...</td>\n      <td>0.0138</td>\n      <td>negative</td>\n    </tr>\n    <tr>\n      <th>2265</th>\n      <td>spoilers ahead it is really unfortunate that a movie so well produced turns out to be such a disappointment i thought this was full of silly clich s and that it basically tried...</td>\n      <td>0.0245</td>\n      <td>negative</td>\n    </tr>\n    <tr>\n      <th>9244</th>\n      <td>no real plot no character development no scorcese level direction but seriously were you really expecting any of this the only thing that matters is that this flick is absolute...</td>\n      <td>0.0294</td>\n      <td>negative</td>\n    </tr>\n    <tr>\n      <th>1645</th>\n      <td>who ever came up with story is one sick person i rented it for our slumber party sleepover and all six of us got freaked out cause we're all in an acting class together and we ...</td>\n      <td>0.0303</td>\n      <td>negative</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                                                                                                                                                                                  excerpt  ...  predicted_sentiment\n9113  honestly i was expecting to hate this one and really only checked it out because jenna jameson is in it but i have to say i got a kick out of evil breed a group of college kids...  ...             negative\n8816  comanche moon had everything going for it for starters simon wincer's back a man who's name is synonymous with high quality tv westerns unfortunately the problems with moon are...  ...             negative\n2265  spoilers ahead it is really unfortunate that a movie so well produced turns out to be such a disappointment i thought this was full of silly clich s and that it basically tried...  ...             negative\n9244  no real plot no character development no scorcese level direction but seriously were you really expecting any of this the only thing that matters is that this flick is absolute...  ...             negative\n1645  who ever came up with story is one sick person i rented it for our slumber party sleepover and all six of us got freaked out cause we're all in an acting class together and we ...  ...             negative\n\n[5 rows x 3 columns]"
     },
     "metadata": {}
    }
   ],
   "source": "# Inspect confident false alerts\nfalse_alert_examples = error_audit[(error_audit['sentiment'] == 'positive') & error_audit['wrong']]\ndisplay(false_alert_examples.sort_values('positive_score')\n        [['excerpt', 'positive_score', 'predicted_sentiment']].head(5).round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-183",
   "metadata": {},
   "source": "**Interpretation guide:** look for mixed opinions, plot summaries, quotations, sarcasm,\nnegation or a discrepancy between an excerpt and the rest of the review. These are\nquestions for manual inspection, not automatically proven explanations of each error.\nPossible label noise should be adjudicated using the original text, not “fixed” to agree\nwith the model. Excerpts are shortened for readability and can omit decisive context.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-086-define-descriptive-length-groups-guide",
   "metadata": {},
   "source": "### 86. Define descriptive length groups\n\nThis is a post-evaluation diagnostic, not a reason to choose a new model on the same test set.\n"
  },
  {
   "cell_type": "code",
   "id": "step-086-define-descriptive-length-groups",
   "metadata": {
    "portfolio_step": 86,
    "purpose": "Define descriptive length groups",
    "execution_scope": "core",
    "execution_seconds": 0.2394
   },
   "execution_count": 86,
   "outputs": [],
   "source": "# Define descriptive length groups\ntest_word_count = test_df['text'].str.split().str.len()\nlength_group = pd.cut(test_word_count, [0, 100, 200, 400, np.inf],\n                      labels=['1–100', '101–200', '201–400', '401+'], include_lowest=True)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-087-compare-errors-across-review-lengths-guide",
   "metadata": {},
   "source": "### 87. Compare errors across review lengths\n\nInclude row counts and class mix. Small slices have more sampling uncertainty; length is not a demographic fairness audit.\n"
  },
  {
   "cell_type": "code",
   "id": "step-087-compare-errors-across-review-lengths",
   "metadata": {
    "portfolio_step": 87,
    "purpose": "Compare errors across review lengths",
    "execution_scope": "core",
    "execution_seconds": 0.1493
   },
   "execution_count": 87,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>words</th>\n      <th>reviews</th>\n      <th>negative_share</th>\n      <th>accuracy</th>\n      <th>macro_f1</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1–100</td>\n      <td>1255</td>\n      <td>0.4486</td>\n      <td>0.9092</td>\n      <td>0.9083</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>101–200</td>\n      <td>4572</td>\n      <td>0.5057</td>\n      <td>0.9103</td>\n      <td>0.9103</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>201–400</td>\n      <td>2774</td>\n      <td>0.5281</td>\n      <td>0.9081</td>\n      <td>0.9080</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>401+</td>\n      <td>1315</td>\n      <td>0.4555</td>\n      <td>0.9125</td>\n      <td>0.9117</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "     words  reviews  negative_share  accuracy  macro_f1\n0    1–100     1255          0.4486    0.9092    0.9083\n1  101–200     4572          0.5057    0.9103    0.9103\n2  201–400     2774          0.5281    0.9081    0.9080\n3     401+     1315          0.4555    0.9125    0.9117"
     },
     "metadata": {}
    }
   ],
   "source": "# Compare errors across review lengths\nslice_rows = []\nfor group in length_group.cat.categories:\n    mask = (length_group == group).to_numpy()\n    if mask.sum() == 0:\n        continue\n    row = metric_summary(str(group), y_test[mask], selected_score[mask])\n    row.update(reviews=int(mask.sum()), negative_share=float((y_test[mask] == 0).mean()))\n    slice_rows.append(row)\nlength_results = pd.DataFrame(slice_rows).rename(columns={'model': 'words'})\ndisplay(length_results[['words', 'reviews', 'negative_share', 'accuracy', 'macro_f1']].round(4))\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-188",
   "metadata": {},
   "source": "### What this evaluation cannot answer\n\n- There is no neutral, mixed or urgent-complaint target; the classifier forces a binary answer.\n- Missing author, film and date fields prevent grouped or future-period evaluation.\n- Exact deduplication does not exclude semantic near duplicates or shared film topics.\n- One split and one seed do not measure full training instability. The bootstrap intervals\n  assume independently sampled reviews and condition on the fitted models.\n- The CSV contains no reliable demographic or language-group audit fields.\n- English movie reviews do not establish transfer performance on business, astronomy or government text.\n- The neural architectures have not been executed here. Their superiority is an open empirical question.\n- This is sentiment classification, not emotion diagnosis, fact checking, spam detection or causal impact analysis.\n\nUseful next experiments are repeated/grouped validation where identifiers are available,\ntemporal evaluation on newly collected data, local language labels, neutral/mixed categories,\nand a correctly separated probability-calibration study. Report any new test set separately.\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-189",
   "metadata": {},
   "source": "## Predict new reviews\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-088-define-prediction-for-new-movie-reviews-guide",
   "metadata": {},
   "source": "### 88. Define prediction for new movie reviews\n\nReuse the training normaliser and vocabulary. Reject empty input rather than returning a confident-looking default.\n"
  },
  {
   "cell_type": "code",
   "id": "step-088-define-prediction-for-new-movie-reviews",
   "metadata": {
    "portfolio_step": 88,
    "purpose": "Define prediction for new movie reviews",
    "execution_scope": "core",
    "execution_seconds": 0.0008
   },
   "execution_count": 88,
   "outputs": [],
   "source": "# Define prediction for new movie reviews\ndef predict_movie_reviews(reviews):\n    texts = [normalise_text(review) for review in reviews]\n    if not texts or any(not text for text in texts):\n        raise ValueError('Supply at least one non-empty review.')\n    if selected_name in core_models:\n        scores = core_models[selected_name].predict_proba(vectorizer.transform(texts))[:, 1]\n    else:\n        tokens = neural_vectorizer(np.asarray(texts, dtype=object)).numpy()\n        scores = neural_models[selected_name].predict(tokens, verbose=0).ravel()\n    return pd.DataFrame({\n        'input': reviews, 'positive_score': scores,\n        'binary_sentiment': np.where(scores >= 0.5, 'positive', 'negative'),\n        'illustrative_negative_flag': (1 - scores) >= triage_threshold,\n    })\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-089-try-clear-negated-mixed-and-neutral-examples-guide",
   "metadata": {},
   "source": "### 89. Try clear, negated, mixed and neutral examples\n\nThese examples illustrate behaviour. They are not an independently labelled accuracy test.\n"
  },
  {
   "cell_type": "code",
   "id": "step-089-try-clear-negated-mixed-and-neutral-examples",
   "metadata": {
    "portfolio_step": 89,
    "purpose": "Try clear, negated, mixed and neutral examples",
    "execution_scope": "core",
    "execution_seconds": 0.0139
   },
   "execution_count": 89,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>input</th>\n      <th>positive_score</th>\n      <th>binary_sentiment</th>\n      <th>illustrative_negative_flag</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>A wonderful film with brilliant acting. I would happily watch it again.</td>\n      <td>0.9924</td>\n      <td>positive</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>A terrible movie. The boring story and poor acting wasted my evening.</td>\n      <td>0.0006</td>\n      <td>negative</td>\n      <td>True</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>This movie was not bad at all; I enjoyed it more than I expected.</td>\n      <td>0.6552</td>\n      <td>positive</td>\n      <td>True</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>I loved the visuals but hated the story.</td>\n      <td>0.8407</td>\n      <td>positive</td>\n      <td>False</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>The movie was released on Friday and runs for two hours.</td>\n      <td>0.2617</td>\n      <td>negative</td>\n      <td>True</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                                                                     input  ...  illustrative_negative_flag\n0  A wonderful film with brilliant acting. I would happily watch it again.  ...                       False\n1    A terrible movie. The boring story and poor acting wasted my evening.  ...                        True\n2        This movie was not bad at all; I enjoyed it more than I expected.  ...                        True\n3                                 I loved the visuals but hated the story.  ...                       False\n4                 The movie was released on Friday and runs for two hours.  ...                        True\n\n[5 rows x 4 columns]"
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Mixed and neutral statements still receive a forced binary label.\nA score is a model estimate, not a verified probability of a human opinion.\n"
    }
   ],
   "source": "# Try clear, negated, mixed and neutral examples\ndemo_reviews = [\n    'A wonderful film with brilliant acting. I would happily watch it again.',\n    'A terrible movie. The boring story and poor acting wasted my evening.',\n    'This movie was not bad at all; I enjoyed it more than I expected.',\n    'I loved the visuals but hated the story.',\n    'The movie was released on Friday and runs for two hours.',\n]\ndisplay(predict_movie_reviews(demo_reviews).round(4))\nprint('Mixed and neutral statements still receive a forced binary label.')\nprint('A score is a model estimate, not a verified probability of a human opinion.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-194",
   "metadata": {},
   "source": "## Business, astronomy and government\n\n**The transferable asset is the analysis workflow.** The IMDb-trained weights have only\nbeen assessed on movie reviews. Each application below is a proposed adaptation requiring\nnew representative text, suitable labels and its own held-out evaluation.\n\n| Perspective | Proposed use | Decision supported | Additional data or model change | Pilot measures |\n|---|---|---|---|---|\n| Business: customer experience | Flag negative product/service feedback for review | Allocate attention and investigate recurring issues | Domain-labelled reviews; neutral/mixed labels; aspect/topic labels | Negative recall, alert precision, queue size, time to review, resolved cases |\n| Business: marketing | Summarise reactions to a campaign | Identify concerns to investigate before a creative change | Campaign/channel/time metadata; representative samples | Theme coverage and human agreement; engagement and conversion evaluated separately |\n| Business: internal audit | Reproduce a text-analysis report | Check exclusions, splits, model versions and error trade-offs | Reviewed labels; documented costs; export provenance | Reconciled row counts, stable split IDs, reproducible outputs, traceable overrides |\n| Astronomy: research literature | Categorise abstracts or observation proposals | Help researchers find relevant work | Astronomy corpus; topic/entity labels; domain-specific representation | Per-topic F1, retrieval recall, expert relevance, minutes saved per search |\n| Astronomy: observatory operations | Route instrument or observation log text | Help staff review technical reports | Labelled operational logs; severity and instrument fields | Routing accuracy, critical-event recall, false alerts, review workload |\n| Government: citizen services | Route and summarise public service feedback | Identify service issues for human review | Local languages; service/location labels; neutral/mixed text; adjudicated samples | Per-language accuracy, critical-issue recall, queue size, response times |\n| Government: consultation analysis | Organise consultation submissions into themes | Help analysts inspect viewpoints and recurring concerns | Multi-label themes; provenance; representative/manual validation | Theme coverage, minority-view retention, human agreement, processing time |\n\n### Business interpretation\n\nSentiment can help prioritise **which feedback to read**. It does not tell a manager which\nproduct defect caused an opinion, whether a customer will leave, or whether a campaign\ncaused sales. Those questions need topic/aspect labels, operational outcomes and a separate\nevaluation design. Auditability is already demonstrated here through data reconciliation,\nsource fingerprinting, split IDs and consistent metrics. A live pilot should compare the\nmodel-assisted workflow with the current manual workflow before claiming an improvement.\n\n### Astronomy interpretation\n\nAn astronomer could adapt text classification to scientific abstracts or observatory logs.\nThe [astroBERT research paper](https://arxiv.org/abs/2112.00590) describes domain-specific NLP\nfor NASA ADS literature discovery and named-entity recognition. It supports the relevance\nof scientific-text modelling; it does not validate this sentiment model for astronomy.\nAstronomical terminology and labels require a new corpus and evaluation. Images, spectra,\nstellar properties and light curves would require different input representations and\ntasks. No astrophysical finding or astronomical classification is produced in this project.\n\n### Government interpretation\n\nThe [OECD’s 2025 account of AI in civic participation](https://www.oecd.org/en/publications/governing-with-artificial-intelligence_795de142-en/full-report/ai-in-civic-participation-and-open-government_51227ce7.html)\ndescribes text analysis, topic clustering and sentiment analysis for consultation inputs.\nThe proposed extension here is to help analysts organise submissions and locate concerns.\nSentiment cannot establish policy correctness or population support: participants and\nonline feedback may be unrepresentative. A South African adaptation needs language coverage,\nneutral/mixed labels and checks that unusual or minority views remain visible. Human\nanalysts should retain the original submissions and responsibility for interpretation.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-090-declare-an-illustrative-business-capacity-scenario-guide",
   "metadata": {},
   "source": "### 90. Declare an illustrative business capacity scenario\n\nThese editable inputs are assumptions for a planning calculation. They are not inferred from the IMDb results or a real business.\n"
  },
  {
   "cell_type": "code",
   "id": "step-090-declare-an-illustrative-business-capacity-scenario",
   "metadata": {
    "portfolio_step": 90,
    "purpose": "Declare an illustrative business capacity scenario",
    "execution_scope": "core",
    "execution_seconds": 0.0002
   },
   "execution_count": 90,
   "outputs": [],
   "source": "# Declare an illustrative business capacity scenario\nMONTHLY_TEXT_ITEMS = 10_000\nMANUAL_MINUTES_PER_ITEM = 2.0\nASSUMED_REVIEW_FRACTION = 0.35\nASSISTED_MINUTES_PER_REVIEW = 1.5\nMONTHLY_OVERSIGHT_HOURS = 15.0\nASSUMED_STAFF_COST_PER_HOUR_ZAR = 180.0\nASSUMED_PLATFORM_COST_PER_MONTH_ZAR = 4_500.0\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-091-calculate-the-assumed-capacity-value-guide",
   "metadata": {},
   "source": "### 91. Calculate the assumed capacity value\n\nFreed staff time is capacity value, not necessarily cash savings. The calculation excludes setup costs and the cost of errors.\n"
  },
  {
   "cell_type": "code",
   "id": "step-091-calculate-the-assumed-capacity-value",
   "metadata": {
    "portfolio_step": 91,
    "purpose": "Calculate the assumed capacity value",
    "execution_scope": "core",
    "execution_seconds": 0.0052
   },
   "execution_count": 91,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>scenario_measure</th>\n      <th>assumed_value</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>Manual hours</td>\n      <td>333.33</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>Assisted + oversight hours</td>\n      <td>102.50</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>Hours freed</td>\n      <td>230.83</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>Assigned capacity value (ZAR)</td>\n      <td>41550.00</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>Platform cost assumption (ZAR)</td>\n      <td>4500.00</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>Capacity value less platform cost (ZAR)</td>\n      <td>37050.00</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                          scenario_measure  assumed_value\n0                             Manual hours         333.33\n1               Assisted + oversight hours         102.50\n2                              Hours freed         230.83\n3            Assigned capacity value (ZAR)       41550.00\n4           Platform cost assumption (ZAR)        4500.00\n5  Capacity value less platform cost (ZAR)       37050.00"
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Illustrative scenario only. Replace every input with measured pilot evidence.\n"
    }
   ],
   "source": "# Calculate the assumed capacity value\nmanual_hours = MONTHLY_TEXT_ITEMS * MANUAL_MINUTES_PER_ITEM / 60\nassisted_hours = (MONTHLY_TEXT_ITEMS * ASSUMED_REVIEW_FRACTION\n                  * ASSISTED_MINUTES_PER_REVIEW / 60 + MONTHLY_OVERSIGHT_HOURS)\nhours_freed = manual_hours - assisted_hours\ncapacity_value = hours_freed * ASSUMED_STAFF_COST_PER_HOUR_ZAR\ndisplay(pd.DataFrame({\n    'scenario_measure': ['Manual hours', 'Assisted + oversight hours', 'Hours freed',\n                         'Assigned capacity value (ZAR)', 'Platform cost assumption (ZAR)',\n                         'Capacity value less platform cost (ZAR)'],\n    'assumed_value': [manual_hours, assisted_hours, hours_freed, capacity_value,\n                      ASSUMED_PLATFORM_COST_PER_MONTH_ZAR,\n                      capacity_value - ASSUMED_PLATFORM_COST_PER_MONTH_ZAR],\n}).round(2))\nprint('Illustrative scenario only. Replace every input with measured pilot evidence.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-092-show-sensitivity-to-the-assumed-review-fraction-guide",
   "metadata": {},
   "source": "### 92. Show sensitivity to the assumed review fraction\n\nMore text requiring human review reduces the capacity benefit. This sensitivity is not a deployment forecast.\n"
  },
  {
   "cell_type": "code",
   "id": "step-092-show-sensitivity-to-the-assumed-review-fraction",
   "metadata": {
    "portfolio_step": 92,
    "purpose": "Show sensitivity to the assumed review fraction",
    "execution_scope": "core",
    "execution_seconds": 0.0051
   },
   "execution_count": 92,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>assumed_review_fraction</th>\n      <th>human_hours_required</th>\n      <th>hours_freed_vs_manual</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0.20</td>\n      <td>65.0</td>\n      <td>268.33</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0.35</td>\n      <td>102.5</td>\n      <td>230.83</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0.50</td>\n      <td>140.0</td>\n      <td>193.33</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>0.75</td>\n      <td>202.5</td>\n      <td>130.83</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1.00</td>\n      <td>265.0</td>\n      <td>68.33</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "   assumed_review_fraction  human_hours_required  hours_freed_vs_manual\n0                     0.20                  65.0                 268.33\n1                     0.35                 102.5                 230.83\n2                     0.50                 140.0                 193.33\n3                     0.75                 202.5                 130.83\n4                     1.00                 265.0                  68.33"
     },
     "metadata": {}
    }
   ],
   "source": "# Show sensitivity to the assumed review fraction\nreview_fractions = np.array([0.20, 0.35, 0.50, 0.75, 1.00])\nscenario_hours = (MONTHLY_TEXT_ITEMS * review_fractions * ASSISTED_MINUTES_PER_REVIEW / 60\n                   + MONTHLY_OVERSIGHT_HOURS)\ndisplay(pd.DataFrame({'assumed_review_fraction': review_fractions,\n                      'human_hours_required': scenario_hours,\n                      'hours_freed_vs_manual': manual_hours - scenario_hours}).round(2))\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-201",
   "metadata": {},
   "source": "**Evidence required for a useful pilot:** collect a representative sample and agree on labels;\ncompare annotation disagreements; reserve future-period data; measure precision, recall and\nworkload; then measure actual review time and outcomes. Set the threshold using organisation\ncosts and available reviewers. Keep model errors and human overrides inspectable. This is a\nconcrete extension plan, not a claim that the IMDb classifier is ready for those domains.\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-202",
   "metadata": {},
   "source": "## Reproducibility and publishing\n\nThe following cells export the evidence behind the page. Split and prediction exports use\nreview IDs rather than review text. The fingerprint records which source CSV they belong\nto. The saved core classifier expects **normalised text**, using the function shown above;\n`predict_core.py` in the package applies that function before prediction.\n\nOnly load joblib/pickle model files from a trusted source. The package’s model artifact is\nan educational snapshot and must be loaded with the recorded scikit-learn version.\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-093-build-the-data-quality-reconciliation-guide",
   "metadata": {},
   "source": "### 93. Build the data-quality reconciliation\n\nThe exclusions must add back to the source row count.\n"
  },
  {
   "cell_type": "code",
   "id": "step-093-build-the-data-quality-reconciliation",
   "metadata": {
    "portfolio_step": 93,
    "purpose": "Build the data-quality reconciliation",
    "execution_scope": "core",
    "execution_seconds": 0.0064
   },
   "execution_count": 93,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>measure</th>\n      <th>count</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>raw_rows</td>\n      <td>50000</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>raw_exact_duplicate_rows</td>\n      <td>418</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>normalised_repeated_rows</td>\n      <td>423</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>quarantined_rows</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>eligible_duplicates_removed</td>\n      <td>423</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>retained_unique_rows</td>\n      <td>49577</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>training_rows</td>\n      <td>29745</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>validation_rows</td>\n      <td>9916</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>test_rows</td>\n      <td>9916</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "                       measure  count\n0                     raw_rows  50000\n1     raw_exact_duplicate_rows    418\n2     normalised_repeated_rows    423\n3             quarantined_rows      0\n4  eligible_duplicates_removed    423\n5         retained_unique_rows  49577\n6                training_rows  29745\n7              validation_rows   9916\n8                    test_rows   9916"
     },
     "metadata": {}
    }
   ],
   "source": "# Build the data-quality reconciliation\nquality_audit = pd.DataFrame({\n    'measure': ['raw_rows', 'raw_exact_duplicate_rows', 'normalised_repeated_rows',\n                'quarantined_rows', 'eligible_duplicates_removed', 'retained_unique_rows',\n                'training_rows', 'validation_rows', 'test_rows'],\n    'count': [len(raw), raw_duplicate_rows, normalised_duplicate_rows, quarantined_rows,\n              duplicates_removed, len(clean), len(train_df), len(val_df), len(test_df)],\n})\ndisplay(quality_audit)\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-094-export-the-score-and-audit-tables-guide",
   "metadata": {},
   "source": "### 94. Export the score and audit tables\n\nThese CSVs support inspection of the exact published run.\n"
  },
  {
   "cell_type": "code",
   "id": "step-094-export-the-score-and-audit-tables",
   "metadata": {
    "portfolio_step": 94,
    "purpose": "Export the score and audit tables",
    "execution_scope": "core",
    "execution_seconds": 0.0074
   },
   "execution_count": 94,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Saved metric, data-quality and diagnostic tables.\n"
    }
   ],
   "source": "# Export the score and audit tables\nquality_audit.to_csv(OUTPUT_DIR / 'data_quality.csv', index=False)\ntuning_results.to_csv(OUTPUT_DIR / 'logistic_validation_search.csv', index=False)\nvalidation_results.to_csv(OUTPUT_DIR / 'validation_metrics.csv', index=False)\ntest_results.to_csv(OUTPUT_DIR / 'test_metrics.csv', index=False)\naccuracy_intervals.to_csv(OUTPUT_DIR / 'accuracy_intervals.csv', index=False)\nthreshold_results.to_csv(OUTPUT_DIR / 'validation_thresholds.csv', index=False)\ntest_triage.to_csv(OUTPUT_DIR / 'test_triage.csv', index=False)\nlength_results.to_csv(OUTPUT_DIR / 'test_length_slices.csv', index=False)\nfeature_weights.to_csv(OUTPUT_DIR / 'logistic_feature_weights.csv', index=False)\nprint('Saved metric, data-quality and diagnostic tables.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-095-export-partition-membership-guide",
   "metadata": {},
   "source": "### 95. Export partition membership\n\nThis allows another analyst with the same normaliser and source CSV to reconstruct the partitions.\n"
  },
  {
   "cell_type": "code",
   "id": "step-095-export-partition-membership",
   "metadata": {
    "portfolio_step": 95,
    "purpose": "Export partition membership",
    "execution_scope": "core",
    "execution_seconds": 0.183
   },
   "execution_count": 95,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Saved 49,577 partition records without review text.\n"
    }
   ],
   "source": "# Export partition membership\nsplit_manifest = pd.concat([\n    frame[['review_id', 'source_row', 'label']].assign(split=name)\n    for name, frame in [('train', train_df), ('validation', val_df), ('test', test_df)]\n], ignore_index=True)\nsplit_manifest.to_csv(OUTPUT_DIR / 'split_manifest.csv', index=False)\nassert split_manifest['review_id'].is_unique\nprint(f'Saved {len(split_manifest):,} partition records without review text.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-096-export-final-selected-predictions-guide",
   "metadata": {},
   "source": "### 96. Export final selected predictions\n\nThe selected score is used consistently in both the classification and illustrative triage policy.\n"
  },
  {
   "cell_type": "code",
   "id": "step-096-export-final-selected-predictions",
   "metadata": {
    "portfolio_step": 96,
    "purpose": "Export final selected predictions",
    "execution_scope": "core",
    "execution_seconds": 0.0668
   },
   "execution_count": 96,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Saved 9,916 final prediction records without review text.\n"
    }
   ],
   "source": "# Export final selected predictions\nprediction_export = test_df[['review_id', 'source_row', 'label']].copy()\nprediction_export['selected_model'] = selected_name\nprediction_export['positive_score'] = selected_score\nprediction_export['predicted_label_default'] = selected_prediction\nprediction_export['negative_flag_locked_policy'] = (1 - selected_score) >= triage_threshold\nprediction_export.to_csv(OUTPUT_DIR / 'selected_test_predictions.csv', index=False)\nprint(f'Saved {len(prediction_export):,} final prediction records without review text.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-097-save-the-strongest-classical-pipeline-components-guide",
   "metadata": {},
   "source": "### 97. Save the strongest classical pipeline components\n\nThis snapshot contains fitted features and classifier together, with an explicit input contract.\n"
  },
  {
   "cell_type": "code",
   "id": "step-097-save-the-strongest-classical-pipeline-components",
   "metadata": {
    "portfolio_step": 97,
    "purpose": "Save the strongest classical pipeline components",
    "execution_scope": "core",
    "execution_seconds": 1.3548
   },
   "execution_count": 97,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Saved sentiment_core.joblib; normalised movie-review text is required.\n"
    }
   ],
   "source": "# Save the strongest classical pipeline components\ncore_snapshot = {\n    'vectorizer': vectorizer, 'classifier': best_core_model,\n    'model_name': best_core_name, 'labels': label_map,\n    'input_contract': 'Apply normalise_text before vectorizer.transform.',\n    'source_sha256': source_sha256,\n    'sklearn_version': importlib.metadata.version('scikit-learn'),\n}\njoblib.dump(core_snapshot, OUTPUT_DIR / 'sentiment_core.joblib', compress=3)\nreloaded_core = joblib.load(OUTPUT_DIR / 'sentiment_core.joblib')\nassert np.allclose(reloaded_core['classifier'].predict_proba(X_val[:5])[:, 1],\n                   validation_scores[best_core_name][:5])\nprint('Saved sentiment_core.joblib; normalised movie-review text is required.')\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-098-save-optional-neural-artifacts-only-after-training-guide",
   "metadata": {},
   "source": "### 98. Save optional neural artifacts only after training\n\nNative Keras weights require the matching vocabulary and sequence length to reproduce inference.\n"
  },
  {
   "cell_type": "code",
   "id": "step-098-save-optional-neural-artifacts-only-after-training",
   "metadata": {
    "portfolio_step": 98,
    "purpose": "Save optional neural artifacts only after training",
    "execution_scope": "optional-neural",
    "execution_seconds": 0.0011
   },
   "execution_count": 98,
   "outputs": [],
   "source": "# Save optional neural artifacts only after training\nif RUN_NEURAL:\n    for name, model in neural_models.items():\n        slug = re.sub(r'[^a-z0-9]+', '_', name.lower()).strip('_')\n        model.save(OUTPUT_DIR / f'{slug}.keras')\n    neural_encoding = {'vocabulary': neural_vectorizer.get_vocabulary(),\n                       'max_length': MAX_LENGTH, 'standardize': None}\n    (OUTPUT_DIR / 'neural_encoding.json').write_text(json.dumps(neural_encoding, indent=2))\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-099-record-the-run-configuration-and-dependency-versions-guide",
   "metadata": {},
   "source": "### 99. Record the run configuration and dependency versions\n\nState what actually ran, and record the uncertainty assumptions rather than hiding them.\n"
  },
  {
   "cell_type": "code",
   "id": "step-099-record-the-run-configuration-and-dependency-versions",
   "metadata": {
    "portfolio_step": 99,
    "purpose": "Record the run configuration and dependency versions",
    "execution_scope": "core",
    "execution_seconds": 0.0306
   },
   "execution_count": 99,
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>version</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>numpy</th>\n      <td>2.3.5</td>\n    </tr>\n    <tr>\n      <th>pandas</th>\n      <td>2.2.3</td>\n    </tr>\n    <tr>\n      <th>scipy</th>\n      <td>1.17.0</td>\n    </tr>\n    <tr>\n      <th>scikit-learn</th>\n      <td>1.8.0</td>\n    </tr>\n    <tr>\n      <th>matplotlib</th>\n      <td>3.10.8</td>\n    </tr>\n    <tr>\n      <th>joblib</th>\n      <td>1.5.3</td>\n    </tr>\n  </tbody>\n</table>\n</div>",
      "text/plain": "             version\nnumpy          2.3.5\npandas         2.2.3\nscipy         1.17.0\nscikit-learn   1.8.0\nmatplotlib    3.10.8\njoblib         1.5.3"
     },
     "metadata": {}
    }
   ],
   "source": "# Record the run configuration and dependency versions\nversions = {name: importlib.metadata.version(name) for name in\n            ['numpy', 'pandas', 'scipy', 'scikit-learn', 'matplotlib', 'joblib']}\nrun_manifest = {\n    'author': 'Gareth Grant', 'dataset_file': DATA_PATH.name, 'dataset_sha256': source_sha256,\n    'python': platform.python_version(), 'dependencies': versions, 'random_state': RANDOM_STATE,\n    'raw_rows': len(raw), 'unique_rows': len(clean),\n    'splits': {'train': len(train_df), 'validation': len(val_df), 'test': len(test_df)},\n    'tfidf_max_features': TFIDF_MAX_FEATURES, 'logistic_candidates_C': LOGISTIC_C_VALUES,\n    'locked_decisions': locked_plan, 'bootstrap_repeats': BOOTSTRAP_REPEATS,\n    'bootstrap_assumption': 'Independent test rows, conditional on fixed trained models.',\n    'neural_requested': RUN_NEURAL, 'neural_models_trained': list(neural_models),\n    'official_stanford_split': False,\n}\nif RUN_NEURAL:\n    run_manifest['tensorflow_version'] = tf.__version__\n(OUTPUT_DIR / 'run_manifest.json').write_text(json.dumps(run_manifest, indent=2))\ndisplay(pd.Series(versions, name='version').to_frame())\n"
  },
  {
   "cell_type": "markdown",
   "id": "step-100-generate-the-portfolio-conclusion-guide",
   "metadata": {},
   "source": "### 100. Generate the portfolio conclusion\n\nThe conclusion changes when the experiment is rerun. It describes evidence and practical limits.\n"
  },
  {
   "cell_type": "code",
   "id": "step-100-generate-the-portfolio-conclusion",
   "metadata": {
    "portfolio_step": 100,
    "purpose": "Generate the portfolio conclusion",
    "execution_scope": "core",
    "execution_seconds": 0.001
   },
   "execution_count": 100,
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Selected on validation: TF-IDF + Logistic Regression.\nFinal accuracy: 90.98%; macro F1: 0.9098.\nConditional bootstrap 95% accuracy interval: 90.44%–91.56%.\nData audit: 50,000 source rows → 49,577 unique valid reviews; 423 eligible duplicates removed.\nThe workflow demonstrates reproducible text analytics, interpretable baselines, error analysis and an explicit operational trade-off.\nBusiness, astronomy and government are adaptation proposals, not validated deployment domains.\nNeural experiments: not run; no performance claim\n"
    }
   ],
   "source": "# Generate the portfolio conclusion\naccuracy_ci = accuracy_intervals.set_index('model').loc[selected_name]\nprint(f'Selected on validation: {selected_name}.')\nprint(f'Final accuracy: {selected_row.accuracy:.2%}; macro F1: {selected_row.macro_f1:.4f}.')\nprint(f'Conditional bootstrap 95% accuracy interval: '\n      f'{accuracy_ci.bootstrap_95_lower:.2%}–{accuracy_ci.bootstrap_95_upper:.2%}.')\nprint(f'Data audit: {len(raw):,} source rows → {len(clean):,} unique valid reviews; '\n      f'{duplicates_removed:,} eligible duplicates removed.')\nprint('The workflow demonstrates reproducible text analytics, interpretable baselines, '\n      'error analysis and an explicit operational trade-off.')\nprint('Business, astronomy and government are adaptation proposals, not validated deployment domains.')\nprint('Neural experiments:', 'trained in this run' if neural_models else 'not run; no performance claim')\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-219",
   "metadata": {},
   "source": "### How to put this on the website\n\nUse the included standalone HTML edition as a reading page, and offer this `.ipynb` as the\ndownloadable notebook. Its page navigation and code toggles let a recruiter read the story\nand a learner inspect individual steps. All charts and styles are embedded; no runtime API\nor model service is needed to view the page. Publishing the HTML does **not** run its Python\nor turn the example predictions into a live predictor.\n\nKeep the HTML and notebook from the same run. After editing or rerunning the notebook,\nregenerate the web preview, for example with `jupyter nbconvert --to html` and your preferred\nstyling. The polished preview supplied here was built from the executed cell outputs.\n\nSuggested portfolio description: “I rebuilt a sentiment analysis workflow to audit text\nquality, prevent duplicate leakage, compare simple and complex model ideas, explain errors\nand assess a decision threshold. The report distinguishes tested results from proposed\nbusiness and public-sector applications.”\n"
  },
  {
   "cell_type": "markdown",
   "id": "note-220",
   "metadata": {},
   "source": "## Practice and references\n\n### Short exercises for learners\n\n1. Explain why the vocabulary must be fitted after the split. Add your answer beside the TF-IDF fit cell.\n2. Remove bigrams, rerun using development data, and explain the effect on negation. Use a new final holdout for a new published comparison.\n3. Explain the top-right confusion-matrix cell in the context of a negative-feedback queue.\n4. Change the assumed missed-negative cost in the development study. Predict how the chosen threshold and queue volume should change before running it.\n5. Read a full misclassified review locally. Suggest a failure category without changing its label to match the model.\n6. Explain why an English movie-review accuracy score cannot establish performance on citizen complaints or astronomy abstracts.\n7. After installing TensorFlow, enable the optional section. Compare validation results, learning curves, parameter counts and training effort before testing a locked choice.\n\n### References\n\n- Maas, A. L., Daly, R. E., Pham, P. T., Huang, D., Ng, A. Y. and Potts, C. (2011).\n  [Learning Word Vectors for Sentiment Analysis](https://aclanthology.org/P11-1015/).\n  ACL, pp. 142–150. [Dataset source and original split](https://ai.stanford.edu/~amaas/data/sentiment/).\n- TensorFlow. [Text classification with an RNN](https://www.tensorflow.org/text/tutorials/text_classification_rnn).\n- Keras. [EarlyStopping](https://keras.io/api/callbacks/early_stopping/).\n- Grezes, F. et al. (2021). [Building astroBERT, a language model for Astronomy & Astrophysics](https://arxiv.org/abs/2112.00590).\n- OECD (2025). [AI in civic participation and open government](https://www.oecd.org/en/publications/governing-with-artificial-intelligence_795de142-en/full-report/ai-in-civic-participation-and-open-government_51227ce7.html),\n  in *Governing with Artificial Intelligence*.\n- Supplied CSV redistribution originally referenced in the source notebooks:\n  [IMDb Dataset of 50K Movie Reviews](https://www.kaggle.com/datasets/lakshmi25npathi/imdb-dataset-of-50k-movie-reviews).\n  The CSV content was audited locally; its original split metadata is not retained.\n\nSources checked 4 October 2026. Dataset interpretations above use the supplied file; external\nsources support the task background and proposed applications. The optional neural code\nuses the modern TensorFlow/Keras API; it remains unexecuted in the supplied edition.\n"
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.12",
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "pygments_lexer": "ipython3"
  },
  "author": "Gareth Grant",
  "portfolio_edition": "1.0",
  "execution_status": "Core executed; optional TensorFlow section unexecuted.",
  "execution_environment": {
   "method": "Sequential Python cell execution with a presentation-only display adapter",
   "core_completed": true,
   "optional_neural_executed": false,
   "code_cells_executed": 100,
   "elapsed_seconds": 57.063
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}