visualize_attention.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import math\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def heat_map(att,color='red'):\n",
    "    words = []\n",
    "    y = []\n",
    "    for a,b in att:\n",
    "        words.append(a)\n",
    "        y.append(b)\n",
    "    y = np.expand_dims(np.array(y), axis=0)\n",
    "    fig, ax = plt.subplots(figsize=(20,0.5))\n",
    "    sns.set(font_scale=1.3)\n",
    "    c = 'red'\n",
    "    if color == 'red':\n",
    "        c = 'OrRd'\n",
    "    elif color == 'blue':\n",
    "        c = 'Blues'\n",
    "    elif color == 'green':\n",
    "        c = 'Greens'\n",
    "    elif color == 'purple':\n",
    "        c = 'BuPu'\n",
    "    sns.heatmap(y,xticklabels=words, cmap=c, linewidths=1.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "tweet = \"rt <@> <number> buff : boston : police and fbi urging anyone with video of the finish line at the time of the explosion should please come forward . .\"\n",
    "attention = [('buff', 0.0254), ('boston', 0.0714), ('police', 0.0928), ('fbi', 0.0732), ('urging', 0.0831), ('anyone', 0.0324), ('with', 0.0359), ('video', 0.023), ('finish', 0.013), ('line', 0.0145), ('time', 0.0187), ('explosion', 0.0384), ('should', 0.0323), ('please', 0.0404), ('come', 0.0306), ('forward', 0.0379)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x36 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "heat_map(attention,'red')"
   ]
  },
  {
   "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}