attention_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "from random import shuffle\n",
    "from src.utils import *\n",
    "import os\n",
    "import sklearn.metrics\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating model with encoder and selector :  pcnn att\n",
      "121 (?, 120, 230)\n",
      "121 (?, 120, 230)\n",
      "(?, 690)\n",
      "Created model with no bootstrapping, bs val : 0.0\n",
      "Setting max class size to :  53\n",
      "Pairs in train dataset NA and non NA :  262336 18252\n",
      "Pairs in train dataset NA and non NA :  94917 1950\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\nall_words, all_pos1, all_pos2, all_masks, all_lengths, all_inst_rels, pair_bag_loc = train_data\\nall_dev_words, all_dev_pos1, all_dev_pos2, all_dev_masks, all_dev_lengths, all_dev_inst_rels, dev_pair_bag_loc = dev_data\\n'"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset_name = \"nyt2\"\n",
    "data_dir = os.path.join(\"data\", dataset_name)\n",
    "encoder = \"pcnn\"\n",
    "selector = \"att\"\n",
    "l2_lambd = 0.0\n",
    "batch_size = 64\n",
    "\n",
    "preprocessed_data_dir = get_preprocessed_dir(data_dir)\n",
    "if not os.path.exists(preprocessed_data_dir):\n",
    "    os.mkdir(preprocessed_data_dir)\n",
    "\n",
    "#preprocessor_batch(data_dir)\n",
    "\n",
    "from src.model import Model\n",
    "tf.reset_default_graph()\n",
    "_, word_vec_mat = load_word_vec(os.path.join(\"data\", dataset_name))\n",
    "max_classes = 53\n",
    "model = Model(word_vec_mat, encoder = encoder, selector=selector, no_of_classes=max_classes)\n",
    "print(\"Setting max class size to : \", max_classes)\n",
    "train_data = load_data(preprocessed_data_dir, max_classes)\n",
    "dev_data = load_data(preprocessed_data_dir, max_classes, \"test\")\n",
    "\n",
    "\n",
    "n_epochs = 4\n",
    "pair_bag_loc = train_data[-1]\n",
    "pairs = list(pair_bag_loc.keys())\n",
    "dev_pair_bag_loc = dev_data[-1]\n",
    "dev_pairs = list(dev_pair_bag_loc.keys())\n",
    "dev_pairs_dict = get_dev_pairs_dict(dev_pairs)\n",
    "n_dev_batches = len(dev_pairs) // batch_size\n",
    "\n",
    "\n",
    "def na_nonNA(pairs):\n",
    "    not_NA_rels = 0\n",
    "    naPairs = []\n",
    "    nonNaPairs = []\n",
    "    for k in pairs:\n",
    "        if k.split(\"#\")[2] != \"0\":\n",
    "            nonNaPairs.append(k)\n",
    "        else:\n",
    "            naPairs.append(k)\n",
    "    return naPairs, nonNaPairs\n",
    "\n",
    "trainNa, trainNonNa = na_nonNA(pairs)\n",
    "devNa, devNonNa = na_nonNA(dev_pairs)\n",
    "not_NA_rels = len(devNonNa)\n",
    "print(\"Pairs in train dataset NA and non NA : \", len(trainNa), len(trainNonNa))\n",
    "print(\"Pairs in train dataset NA and non NA : \", len(devNa), len(devNonNa))\n",
    "\"\"\"\n",
    "all_words, all_pos1, all_pos2, all_masks, all_lengths, \\\n",
    "all_inst_rels, pair_bag_loc = train_data\n",
    "all_dev_words, all_dev_pos1, all_dev_pos2, all_dev_masks, \\\n",
    "all_dev_lengths, all_dev_inst_rels, dev_pair_bag_loc = dev_data\n",
    "\"\"\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "path : saved/models/pcnn_att_nyt2_none_53_n_0_0.3455\n",
      "INFO:tensorflow:Restoring parameters from saved/models/pcnn_att_nyt2_none_53_n_0_0.3455\n"
     ]
    }
   ],
   "source": [
    "model.mloader(os.path.join(\"saved\", \"models\", \"pcnn_att_nyt2_none_53_n_0_0.3455\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test_dev(dev_pairs = dev_pairs):\n",
    "    test_res = []\n",
    "    for i in range(n_dev_batches):\n",
    "      batch_keys = dev_pairs[i * batch_size : (i + 1) * batch_size]\n",
    "      words, pos1, pos2, inst_rels, masks, lengths, \\\n",
    "        rels, scope = batch_maker(dev_data, batch_keys)\n",
    "      pos1[pos1 > 239] = 239\n",
    "      pos2[pos2 > 239] = 239\n",
    "      pos1[pos1 < 0] = 0\n",
    "      pos2[pos2 < 0] = 0\n",
    "      output, atts = model.test_batch(words, pos1, pos2, inst_rels, \n",
    "        masks, lengths, rels, scope)\n",
    "      for i, k in enumerate(batch_keys):\n",
    "        entPair = \"#\".join(k.split(\"#\")[:2])\n",
    "        entPairRels = dev_pairs_dict[entPair]\n",
    "        for j in range(1, 53):\n",
    "          correct = 0\n",
    "          if j in entPairRels:\n",
    "            correct = 1\n",
    "          if output[i][j] > 0.01:\n",
    "            test_res.append({\"entPair\" : entPair, \n",
    "                \"score\" : output[i][j], \n",
    "                \"correct\" : correct, \n",
    "                \"pred\" : j, \n",
    "                \"actual\" : entPairRels,\n",
    "                \"atts\" : atts[i][j]})\n",
    "#       print(test_res)\n",
    "    prec = []\n",
    "    recall = []\n",
    "    correct = 0\n",
    "    sorted_test_result = sorted(test_res, key=lambda x: x['score'], reverse = True)\n",
    "    \n",
    "    import time\n",
    "    start = time.time()\n",
    "\n",
    "    for i, item in enumerate(sorted_test_result):\n",
    "      if item[\"correct\"]:\n",
    "        correct += 1  \n",
    "      prec.append(float(correct) / (i + 1))\n",
    "      recall.append(float(correct) / not_NA_rels)\n",
    "    return prec, recall, sorted_test_result[:5000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "p, r, sorted_test_result = test_dev()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "use_top = 100\n",
    "\n",
    "atts_max = []\n",
    "atts_min = []\n",
    "bag_lengths = []\n",
    "atts_max_filter_3_instances = []\n",
    "atts_max_2nd_filter_3_instances = []\n",
    "atts_min_filter_3_instances = []\n",
    "bag_lengths_filter_3_instances = []\n",
    "for i in range(use_top):\n",
    "    atts_max.append(max(sorted_test_result[i]['atts']))\n",
    "    bag_lengths.append(len(sorted_test_result[i]['atts']))\n",
    "    atts_min.append(min(sorted_test_result[i]['atts']))\n",
    "    if len(sorted_test_result[i]['atts']) > 2:\n",
    "        atts_max_filter_3_instances.append(max(sorted_test_result[i]['atts']))\n",
    "        atts_max_2nd_filter_3_instances.append(sorted(sorted_test_result[i]['atts'], reverse = True)[1])\n",
    "        atts_min_filter_3_instances.append(min(sorted_test_result[i]['atts']))\n",
    "        bag_lengths_filter_3_instances.append(len(sorted_test_result[i]['atts']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "fig, axs = plt.subplots(1,2,figsize=(15,10))\n",
    "ax_iter = iter(axs.flat)\n",
    "plots = [atts_max, atts_min,\n",
    "         bag_lengths, bag_lengths_filter_3_instances,\n",
    "         atts_max, atts_max_filter_3_instances,\n",
    "         atts_max_filter_3_instances, atts_min_filter_3_instances,\n",
    "         atts_max_filter_3_instances, atts_max_2nd_filter_3_instances]\n",
    "ax = next(ax_iter)\n",
    "sns.distplot(atts_max, ax=ax, hist_kws=dict(cumulative=True), kde_kws=dict(cumulative=True), axlabel=\"maximum attention in bag\")\n",
    "ax = next(ax_iter)\n",
    "sns.distplot(atts_max_filter_3_instances, ax=ax, hist_kws=dict(cumulative=True), kde_kws=dict(cumulative=True), axlabel=\"maximum attention in bag after filter\")\n",
    "plt.savefig(\"Attention_weights_dist.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axs = plt.subplots(1,2,figsize=(15,10))\n",
    "ax_iter = iter(axs.flat)\n",
    "ax = next(ax_iter)\n",
    "sns.distplot(bag_lengths, bins = [i for i in range(14)], ax=ax, hist_kws=dict(cumulative=True), kde_kws=dict(cumulative=True), axlabel=\"no of sentences in bag\")\n",
    "ax = next(ax_iter)\n",
    "sns.distplot(bag_lengths_filter_3_instances, bins = [i for i in range(3, 14)], ax=ax, hist_kws=dict(cumulative=True), kde_kws=dict(cumulative=True), axlabel=\"no of sentences in bag after filter\")\n",
    "plt.savefig(\"Bag_lengths_dist.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[7, 1, 1, 1, 1, 1, 1, 1, 1, 1, 13, 1, 1, 1, 4, 2, 1, 1, 3, 2, 5, 1, 2, 2, 1, 2, 1, 1, 1, 1, 1, 1, 3, 1, 1, 1, 2, 1, 1, 2, 1, 3, 2, 1, 1, 1, 3, 1, 3, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 4, 3, 7, 7, 1, 2, 1, 1, 3, 4, 1, 1, 2, 6, 1, 1, 2, 3, 3, 3, 4, 2, 1, 2, 1, 7, 2, 1, 1, 3, 3, 1, 1, 1, 1, 1, 1, 2, 1, 10, 1]\n",
      "Counter({1: 60, 2: 16, 3: 12, 7: 4, 4: 4, 13: 1, 5: 1, 6: 1, 10: 1})\n"
     ]
    }
   ],
   "source": [
    "from collections import Counter\n",
    "print(bag_lengths)\n",
    "print(Counter(bag_lengths))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AxesSubplot(0.125,0.125;0.775x0.755)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.ylim(0.13, 0.19)\n",
    "temp = [(50 , 0.1806), (100, 0.1748), (150, 0.1531), (200, 0.1564), (250, 0.1550), (300, 0.1554), (350, 0.1506), (400, 0.1392), (450, 0.1360), (503, 0.1379)]\n",
    "x = [t[0] for t in temp]\n",
    "y = [t[1] for t in temp]\n",
    "fig = sns.barplot(x, y)\n",
    "fig.set(xlabel=\"no of classes\", ylabel=\"AUC score\")\n",
    "print(fig)\n",
    "fig = fig.figure\n",
    "fig.savefig(\"temp\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}