annotation_artifacts.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Annotation Artifacts in Natural Language Inference Data\n",
    "\n",
    "This notebook analyses annotations in the [SNLI](https://nlp.stanford.edu/projects/snli/) and [MultiNLI](https://www.nyu.edu/projects/bowman/multinli/) datasets, showing consistent patterns in the data which might cause models to achieve inflated performance.\n",
    "\n",
    "If using in your research, please cite [our NAACL paper](https://aclweb.org/anthology/N18-2017)\n",
    "\n",
    "```\n",
    "@inproceedings{gururangan2018annotation,\n",
    "  title={Annotation Artifacts in Natural Language Inference Data},\n",
    "  author={Gururangan, Suchin and Swayamdipta, Swabha and Levy, Omer and Schwartz, Roy and Bowman, Samuel and Smith, Noah A},\n",
    "  booktitle={Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers)},\n",
    "  pages={107--112},\n",
    "  year={2018}\n",
    "}\n",
    "```\n",
    "\n",
    "## Installation\n",
    "\n",
    "First, you would need to install some packages:\n",
    "\n",
    "```\n",
    "pip install matplotlib\n",
    "pip install nltk\n",
    "pip install numpy\n",
    "pip install pandas\n",
    "pip install re\n",
    "pip install seaborn\n",
    "pip install string\n",
    "pip install tqdm\n",
    "```\n",
    "\n",
    "## Load dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import re\n",
    "import seaborn as sns\n",
    "import string\n",
    "\n",
    "from collections import Counter\n",
    "from nltk import PorterStemmer\n",
    "from IPython.core.debugger import Tracer\n",
    "from tqdm import tqdm\n",
    "\n",
    "#tqdm.pandas(\"mybar!\")\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load Data\n",
    "\n",
    "Here, we only show SNLI, but MNLI, being in the same format can be processed similarly."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "%%capture\n",
    "\n",
    "snli = pd.read_csv('/Users/swabhas/data/aflite/snli_1.0_glue/train.tsv',\n",
    "                   sep = '\\t',\n",
    "                   error_bad_lines=False).dropna(subset=['sentence2',\n",
    "                                                         'sentence2_binary_parse',\n",
    "                                                         'sentence2_parse'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "def tokenize(string):\n",
    "    \"\"\"\n",
    "    tokenize string from binary parse\n",
    "    \"\"\"\n",
    "    string = re.sub(r'\\(|\\)', '', str(string))\n",
    "    return string.lower().split()\n",
    "\n",
    "def plot_tokens(df):\n",
    "    \"\"\"\n",
    "    plot token length dist\n",
    "    \"\"\"\n",
    "    total_count = df.sentence2.apply(lambda x: len(x)).sum()\n",
    "    ax = (df.sentence2_binary_parse\n",
    "            .apply(lambda x: len(tokenize(x)))\n",
    "            .value_counts()\n",
    "            .sort_index()\n",
    "            .div(total_count)\n",
    "            .plot(kind='line', legend=True, ylim=[0 , 0.006]))\n",
    "    return ax\n",
    "\n",
    "def plot_characters(df):\n",
    "    \"\"\"\n",
    "    plot character length dist\n",
    "    \"\"\"\n",
    "    total_count = df.sentence2.apply(lambda x: len(x)).sum()\n",
    "    ax = (df.sentence2\n",
    "            .apply(lambda x: len(x))\n",
    "            .value_counts()\n",
    "            .sort_index()\n",
    "            .div(total_count)\n",
    "            .plot(kind='line', legend=True, ylim=[0 , 0.0012]))\n",
    "    return ax\n",
    "\n",
    "def ppmi(df):\n",
    "    \"\"\"\n",
    "    computes positive pointwise mutual information between labels and words in the hypothesis.\n",
    "    Args:\n",
    "        df -> dataframe (mnli or snli)\n",
    "    Returns:\n",
    "        ppmis -> pmi data\n",
    "    \"\"\"\n",
    "    alpha = 10.0  # for smoothing\n",
    "    \n",
    "    # get joint (w,c) counts\n",
    "    Px = []\n",
    "    Py = []\n",
    "    Pxy = []\n",
    "    for idx, row in df.iterrows():\n",
    "        for word in tokenize(row.sentence2_binary_parse):\n",
    "            Pxy.append((word, row.gold_label))\n",
    "            Px.append(word)\n",
    "        Py.append(row.gold_label)\n",
    "    Pxy = Counter(Pxy)\n",
    "    Px = Counter(Px)\n",
    "    Py = Counter(Py)\n",
    "    Py = {y: c + len(Px)*alpha for y, c in Py.items()}\n",
    "    Px = {x: c + 3*alpha for x, c in Px.items()}\n",
    "    total_word_count = sum(Px.values())\n",
    "\n",
    "    # get c counts\n",
    "    ppmis = []\n",
    "    \n",
    "    seen_pairs = {}\n",
    "    total_Py = 3\n",
    "    for word in Px.keys():\n",
    "        for cls in ['contradiction', 'entailment', 'neutral']:\n",
    "            if seen_pairs.get((word, cls)) is None:\n",
    "                # artificial smoothing\n",
    "                word_count = float(Px[word])\n",
    "                cls_count = float(Py[cls])\n",
    "                word_cls_count = Pxy.get((word, cls), 0.0) + alpha\n",
    "                z = (float(word_cls_count) * total_word_count / (float(word_count) * float(cls_count)))\n",
    "                # ppmi\n",
    "                ppmi = max([np.log(z), 0.0])\n",
    "                output = {'word': word,\n",
    "                          'class': cls,\n",
    "                          'ppmi': ppmi,\n",
    "                          'count': word_count - 3*alpha,\n",
    "                          'wc_count': Pxy.get((word, cls), 0.0),}\n",
    "                seen_pairs[(word, cls)] = 1\n",
    "                ppmis.append(output)\n",
    "    return pd.DataFrame(ppmis)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PPMI and Token Length Distributions\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "snli_ppmi = ppmi(snli)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Top 20 PPMIs across (class, word) pairs in SNLI"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "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>class</th>\n",
       "      <th>count</th>\n",
       "      <th>ppmi</th>\n",
       "      <th>wc_count</th>\n",
       "      <th>word</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>651</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>2192.0</td>\n",
       "      <td>2.359585</td>\n",
       "      <td>2182.0</td>\n",
       "      <td>nobody</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4377</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>759.0</td>\n",
       "      <td>2.280274</td>\n",
       "      <td>709.0</td>\n",
       "      <td>cats</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3191</th>\n",
       "      <td>neutral</td>\n",
       "      <td>339.0</td>\n",
       "      <td>2.267198</td>\n",
       "      <td>322.0</td>\n",
       "      <td>championship</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5636</th>\n",
       "      <td>neutral</td>\n",
       "      <td>159.0</td>\n",
       "      <td>2.261011</td>\n",
       "      <td>159.0</td>\n",
       "      <td>joyously</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4146</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>534.0</td>\n",
       "      <td>2.260701</td>\n",
       "      <td>494.0</td>\n",
       "      <td>nothing</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>167</th>\n",
       "      <td>neutral</td>\n",
       "      <td>594.0</td>\n",
       "      <td>2.248444</td>\n",
       "      <td>541.0</td>\n",
       "      <td>vacation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4681</th>\n",
       "      <td>entailment</td>\n",
       "      <td>364.0</td>\n",
       "      <td>2.245740</td>\n",
       "      <td>337.0</td>\n",
       "      <td>least</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>243.0</td>\n",
       "      <td>2.244346</td>\n",
       "      <td>230.0</td>\n",
       "      <td>frowning</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1497</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>949.0</td>\n",
       "      <td>2.242416</td>\n",
       "      <td>849.0</td>\n",
       "      <td>naked</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>347</th>\n",
       "      <td>neutral</td>\n",
       "      <td>640.0</td>\n",
       "      <td>2.235483</td>\n",
       "      <td>574.0</td>\n",
       "      <td>winning</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1727</th>\n",
       "      <td>neutral</td>\n",
       "      <td>1939.0</td>\n",
       "      <td>2.234164</td>\n",
       "      <td>1704.0</td>\n",
       "      <td>tall</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10112</th>\n",
       "      <td>neutral</td>\n",
       "      <td>275.0</td>\n",
       "      <td>2.232278</td>\n",
       "      <td>255.0</td>\n",
       "      <td>siblings</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>713</th>\n",
       "      <td>neutral</td>\n",
       "      <td>681.0</td>\n",
       "      <td>2.231056</td>\n",
       "      <td>607.0</td>\n",
       "      <td>favorite</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4202</th>\n",
       "      <td>neutral</td>\n",
       "      <td>158.0</td>\n",
       "      <td>2.230168</td>\n",
       "      <td>153.0</td>\n",
       "      <td>impress</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>168</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>6390.0</td>\n",
       "      <td>2.221051</td>\n",
       "      <td>5504.0</td>\n",
       "      <td>sleeping</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3543</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>139.0</td>\n",
       "      <td>2.220014</td>\n",
       "      <td>135.0</td>\n",
       "      <td>mars</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1971</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>1581.0</td>\n",
       "      <td>2.219855</td>\n",
       "      <td>1372.0</td>\n",
       "      <td>tv</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2061</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>299.0</td>\n",
       "      <td>2.219028</td>\n",
       "      <td>272.0</td>\n",
       "      <td>anything</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8198</th>\n",
       "      <td>neutral</td>\n",
       "      <td>530.0</td>\n",
       "      <td>2.216624</td>\n",
       "      <td>469.0</td>\n",
       "      <td>date</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6957</th>\n",
       "      <td>contradiction</td>\n",
       "      <td>303.0</td>\n",
       "      <td>2.210483</td>\n",
       "      <td>273.0</td>\n",
       "      <td>quietly</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               class   count      ppmi  wc_count          word\n",
       "651    contradiction  2192.0  2.359585    2182.0        nobody\n",
       "4377   contradiction   759.0  2.280274     709.0          cats\n",
       "3191         neutral   339.0  2.267198     322.0  championship\n",
       "5636         neutral   159.0  2.261011     159.0      joyously\n",
       "4146   contradiction   534.0  2.260701     494.0       nothing\n",
       "167          neutral   594.0  2.248444     541.0      vacation\n",
       "4681      entailment   364.0  2.245740     337.0         least\n",
       "81     contradiction   243.0  2.244346     230.0      frowning\n",
       "1497   contradiction   949.0  2.242416     849.0         naked\n",
       "347          neutral   640.0  2.235483     574.0       winning\n",
       "1727         neutral  1939.0  2.234164    1704.0          tall\n",
       "10112        neutral   275.0  2.232278     255.0      siblings\n",
       "713          neutral   681.0  2.231056     607.0      favorite\n",
       "4202         neutral   158.0  2.230168     153.0       impress\n",
       "168    contradiction  6390.0  2.221051    5504.0      sleeping\n",
       "3543   contradiction   139.0  2.220014     135.0          mars\n",
       "1971   contradiction  1581.0  2.219855    1372.0            tv\n",
       "2061   contradiction   299.0  2.219028     272.0      anything\n",
       "8198         neutral   530.0  2.216624     469.0          date\n",
       "6957   contradiction   303.0  2.210483     273.0       quietly"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "snli_ppmi.sort_values(by='ppmi', ascending=False).head(n=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Graphs for PPMI"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Top 20 words occuring with the class `contradiction`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set_style(\"whitegrid\", {'axes.grid' : False})\n",
    "ax = (snli_ppmi.loc[snli_ppmi['class'] == 'contradiction']\n",
    "              .sort_values(by='ppmi', ascending=False)\n",
    "              .head(n=20)\n",
    "              .plot(kind='barh', x='word', y='ppmi', title='top 20 PPMI(word, contradiction) SNLI', legend=False))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Top 20 words occuring with the class `entailment`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set_style(\"whitegrid\", {'axes.grid' : False})\n",
    "ax = (snli_ppmi.loc[snli_ppmi['class'] == 'entailment']\n",
    "              .sort_values(by='ppmi', ascending=False)\n",
    "              .head(n=20)\n",
    "              .plot(kind='barh', x='word', y='ppmi', title='top 20 PPMI(word, entailment) SNLI', legend=False))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Top 20 words occuring with the class `neutral`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set_style(\"whitegrid\", {'axes.grid' : False})\n",
    "ax = (snli_ppmi.loc[snli_ppmi['class'] == 'neutral']\n",
    "              .sort_values(by='ppmi', ascending=False)\n",
    "              .head(n=20)\n",
    "              .plot(kind='barh', x='word', y='ppmi', title='top 20 PPMI(word, neutral) SNLI', legend=False))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Token Length Distributions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "contradiction_snli = snli.loc[snli.gold_label == 'contradiction']\n",
    "entailment_snli = snli.loc[snli.gold_label == 'entailment']\n",
    "neutral_snli = snli.loc[snli.gold_label == 'neutral']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1289dfc50>"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "_ = plot_tokens(neutral_snli)\n",
    "_ = plot_tokens(contradiction_snli)\n",
    "ax = plot_tokens(entailment_snli)\n",
    "ax.set_xlabel('length of hypothesis sentence')\n",
    "ax.set_ylabel('density')\n",
    "ax.legend([\"snli neutral\",\n",
    "           \"snli contradiction\",\n",
    "           \"snli entailment\"])\n",
    "# ax.set_ylim([0, 0.0012])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}