tutorials/explainers/LIME/lime_text.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "6f6affbc",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "<img width=\"150\" alt=\"Logo_ER10\" src=\"https://user-images.githubusercontent.com/3244249/151994514-b584b984-a148-4ade-80ee-0f88b0aefa45.png\">\n",
    "\n",
    "### Interpreting a movie review sentiment model with LIME\n",
    "This notebook demonstrates the use of DIANNA with the LIME method on the [Stanford Sentiment Treebank dataset](https://nlp.stanford.edu/sentiment/index.html) which contains one-sentence movie reviews. See also [their paper](https://nlp.stanford.edu/~socherr/EMNLP2013_RNTN.pdf). A pre-trained neural network classifier is used, which identifies whether a movie review is positive or negative.\n",
    "\n",
    "LIME (Local Interpretable Model-agnostic Explanations) is an explainable-AI method that aims to create an interpretable model that locally represents the classifier. For more details see the [LIME paper](https://arxiv.org/abs/1602.04938)."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "224e1862",
   "metadata": {},
   "source": [
    "#### Colab Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "df22c5aa",
   "metadata": {},
   "outputs": [],
   "source": [
    "running_in_colab = 'google.colab' in str(get_ipython())\n",
    "if running_in_colab:\n",
    "    # install dianna\n",
    "    !python3 -m pip install dianna[notebooks]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "091947e1",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "#### 1. Imports and paths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bf888085",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import spacy\n",
    "from torchtext.vocab import Vectors\n",
    "from scipy.special import expit as sigmoid\n",
    "\n",
    "import dianna\n",
    "from dianna import visualization\n",
    "from dianna import utils\n",
    "from dianna.utils.downloader import download\n",
    "from dianna.utils.tokenizers import SpacyTokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "4263ce5c",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "model_path = download('movie_review_model.onnx', 'model')\n",
    "word_vector_path = download('movie_reviews_word_vectors.txt', 'data')\n",
    "labels = (\"negative\", \"positive\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "86fd772b",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "#### 2. Loading the model\n",
    "\n",
    "The classifier is stored in ONNX format. It accepts numerical tokens as input, and outputs a score between 0 and 1, where 0 means the review is negative and 1 that it is positive.  \n",
    "Here we define a class to run the model, which accepts a sentence (i.e. string) as input instead and returns two classes: negative and positive."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f09c7fc3",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ensure the tokenizer for english is available\n",
    "spacy.cli.download('en_core_web_sm')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "02e6d4b1",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "class MovieReviewsModelRunner:\n",
    "    def __init__(self, model, word_vectors, max_filter_size):\n",
    "        self.run_model = utils.get_function(model)\n",
    "        self.vocab = Vectors(word_vectors, cache=os.path.dirname(word_vectors))\n",
    "        self.max_filter_size = max_filter_size\n",
    "        \n",
    "        self.tokenizer = SpacyTokenizer(name='en_core_web_sm')\n",
    "\n",
    "    def __call__(self, sentences):\n",
    "        # ensure the input has a batch axis\n",
    "        if isinstance(sentences, str):\n",
    "            sentences = [sentences]\n",
    "\n",
    "        output = []\n",
    "        for sentence in sentences:\n",
    "            # tokenize and pad to minimum length\n",
    "            tokens = self.tokenizer.tokenize(sentence.lower())\n",
    "            if len(tokens) < self.max_filter_size:\n",
    "                tokens += ['<pad>'] * (self.max_filter_size - len(tokens))\n",
    "            \n",
    "            # numericalize the tokens\n",
    "            tokens_numerical = [self.vocab.stoi[token] if token in self.vocab.stoi else self.vocab.stoi['<unk>']\n",
    "                                for token in tokens]\n",
    "\n",
    "            # run the model, applying a sigmoid because the model outputs logits, remove any remaining batch axis\n",
    "            pred = float(sigmoid(self.run_model([tokens_numerical])))\n",
    "            output.append(pred)\n",
    "\n",
    "        # output two classes\n",
    "        positivity = np.array(output)\n",
    "        negativity = 1 - positivity\n",
    "        return np.transpose([negativity, positivity])       "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "214007db",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▉| 13888/13889 [00:00<00:00, 29837.09it/s]\n"
     ]
    }
   ],
   "source": [
    "# define model runner. max_filter_size is a property of the model\n",
    "model_runner = MovieReviewsModelRunner(model_path, word_vector_path, max_filter_size=5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7939aef9",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "#### 3. Applying LIME with DIANNA\n",
    "The simplest way to run DIANNA on text data is with `dianna.explain_text`. The arguments are:\n",
    "* The function that runs the model (a path to a model in ONNX format is also accepted)\n",
    "* The text we want to explain\n",
    "* The name of the explainable-AI method we want to use, here LIME\n",
    "* The numerical index of the class we want an explanation for\n",
    "\n",
    "`dianna.explain_text` returns a list of tuples. Each tuple contains a word, its location in the input text, and its importance for the selected output class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "083ff6b1",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "review = \"A delectable and intriguing thriller filled with surprises.\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "1b2b2ab2",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/rj/_8tmcy8118q5h1zc_3dz1rz00000gn/T/ipykernel_32287/1753736335.py:26: DeprecationWarning: Conversion of an array with ndim > 0 to a scalar is deprecated, and will error in future. Ensure you extract a single element from your array before performing this operation. (Deprecated NumPy 1.25.)\n",
      "  pred = float(sigmoid(self.run_model([tokens_numerical])))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('intriguing', 3, 0.1494251390016417),\n",
       " ('thriller', 4, 0.06169666393710174),\n",
       " ('delectable', 1, 0.05673652647678169),\n",
       " ('A', 0, 0.01974513430826506),\n",
       " ('and', 2, 0.01822693342907461),\n",
       " ('with', 6, 0.014052819256237513),\n",
       " ('filled', 5, -0.011330645775131239),\n",
       " ('.', 8, 0.0037673252252354216),\n",
       " ('surprises', 7, 0.0035828865798261094)]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# We're getting the explanation for the 'positive' class only,\n",
    "# but dianna supports explaining for multiple labels in one go.\n",
    "# It therefore always outputs a list of saliency maps. We want\n",
    "# the first and only saliency map from this list here.\n",
    "explanation_relevance = dianna.explain_text(model_runner, review, model_runner.tokenizer,\n",
    "                                            'LIME', labels=[labels.index('positive')])[0]\n",
    "explanation_relevance"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3ef466a",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "#### 4. Visualization\n",
    "DIANNA includes a visualization package, capable of highlighting the relevance of each word in the text for a chosen class. The visualization is in HTML format.\n",
    "Words in favour of the selected class are highlighted in red, while words against the selected class - in blue."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "1bbba305",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x72 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, _ = visualization.highlight_text(explanation_relevance, model_runner.tokenizer.tokenize(review))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1f460858",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "The most important word for this review being classified as positive is \"intriguing\", followed by \"delectable\" and \"thriller\".\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9818c180-322d-4a52-8170-7f7e0fe70e31",
   "metadata": {},
   "source": [
    "#### Trying more text explanations"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d997d9c0-6277-4778-b9a4-3248a8c8ea93",
   "metadata": {},
   "source": [
    "And what are the words' relevances to a '**negative**' classification decision?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "24ffddee-6e39-4a0d-876d-2622c96a5195",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/rj/_8tmcy8118q5h1zc_3dz1rz00000gn/T/ipykernel_32287/1753736335.py:26: DeprecationWarning: Conversion of an array with ndim > 0 to a scalar is deprecated, and will error in future. Ensure you extract a single element from your array before performing this operation. (Deprecated NumPy 1.25.)\n",
      "  pred = float(sigmoid(self.run_model([tokens_numerical])))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('intriguing', 3, -0.15120818911158293),\n",
       " ('thriller', 4, -0.06503445215836798),\n",
       " ('delectable', 1, -0.05959389939279887),\n",
       " ('A', 0, -0.01926809327783268),\n",
       " ('and', 2, -0.014870616516598822),\n",
       " ('with', 6, -0.012206798269964911),\n",
       " ('filled', 5, 0.00912963811840552),\n",
       " ('surprises', 7, -0.006663047358285325),\n",
       " ('.', 8, -0.001571926230751491)]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "explanation_relevance = dianna.explain_text(model_runner, review, model_runner.tokenizer,\n",
    "                                            'LIME', labels=[labels.index('negative')])[0]\n",
    "explanation_relevance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "9043f3e5-79da-4dbf-8190-17b6ecdf2f1d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x72 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, _ = visualization.highlight_text(explanation_relevance, model_runner.tokenizer.tokenize(review))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d27ae57f-f2ff-4089-b804-06e58ce5af67",
   "metadata": {},
   "source": [
    "And what about a harder to classify review with mixed words, but still a clear overall sentiment?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e3846fef-abfd-4e49-b4de-bf437205dd18",
   "metadata": {},
   "outputs": [],
   "source": [
    "mixed_review = 'The movie started great but the ending is boring and unoriginal!'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "919f2f12-7c04-4fa2-9f83-2043a450dc07",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/rj/_8tmcy8118q5h1zc_3dz1rz00000gn/T/ipykernel_32287/1753736335.py:26: DeprecationWarning: Conversion of an array with ndim > 0 to a scalar is deprecated, and will error in future. Ensure you extract a single element from your array before performing this operation. (Deprecated NumPy 1.25.)\n",
      "  pred = float(sigmoid(self.run_model([tokens_numerical])))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('boring', 8, 0.4518653570721131),\n",
       " ('great', 3, -0.22827878280499303),\n",
       " ('ending', 6, 0.11350805065432693),\n",
       " ('!', 11, 0.0957722037393977),\n",
       " ('unoriginal', 10, 0.052681002368695704),\n",
       " ('movie', 1, -0.04113332032878819),\n",
       " ('is', 7, 0.034230793363654734),\n",
       " ('the', 5, -0.022376073721346687),\n",
       " ('and', 9, 0.013376920413792116),\n",
       " ('but', 4, 0.009010962798098054)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "explanation_relevance = dianna.explain_text(model_runner, mixed_review, model_runner.tokenizer,\n",
    "                                            'LIME', labels=[labels.index('negative')])[0]\n",
    "explanation_relevance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "86247d24-6c7e-46d5-89c0-2f51c7905988",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x72 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, _ = visualization.highlight_text(explanation_relevance, model_runner.tokenizer.tokenize(mixed_review))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3ad9c13-cfd2-4cf7-a03f-322326839406",
   "metadata": {},
   "source": [
    "Indeed, \"boring\" and \"unoriginal\" are the words carrying the overall negative sentiment, hence they contribute most toward that decision, while \"great\" being a positive word, contributes negatively towards the classifier's decision!"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}