D0-w2vModel.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training Vectorizing Words using Word2Vec model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import gensim, logging\n",
    "logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)\n",
    "import multiprocessing as mp\n",
    "import os\n",
    "import pandas as pd\n",
    "import cPickle\n",
    "import numpy as np\n",
    "from sklearn.decomposition import PCA\n",
    "import matplotlib.font_manager\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "#import visword2vec\n",
    "from collections import Counter\n",
    "\n",
    "core_count= mp.cpu_count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def plot1( words,vectors):\n",
    "\n",
    "    # do PCA\n",
    "    #X = np.mat[vectors]\n",
    "    pca = PCA(n_components=2)\n",
    "    X=vectors\n",
    "    pca.fit(X)\n",
    "    print pca.explained_variance_ratio_\n",
    "    X = pca.transform(X)\n",
    "    xs = X[:, 0]\n",
    "    ys = X[:, 1]\n",
    "\n",
    "    # draw\n",
    "    plt.figure(figsize=(6,6))\n",
    "    plt.scatter(xs, ys, marker = 'o',s=5)\n",
    "    for i, w in enumerate(words):\n",
    "        plt.annotate(\n",
    "            w.decode('utf-8', 'ignore'),\n",
    "            xy = (xs[i], ys[i]), xytext = (3, 3),\n",
    "            textcoords = 'offset points', ha = 'left', va = 'top')\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training word2vec model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The output already exists!\n"
     ]
    }
   ],
   "source": [
    "if os.path.isfile('../DataFiles/trained_vec.bin'):\n",
    "    print 'The output already exists!'\n",
    "else:\n",
    "    train_file = open('../DataFiles/all_reviews_sentences.txt')\n",
    "    sentences= []\n",
    "    for line in train_file:\n",
    "        sentences.append(line.lower().replace('.','').replace('-','').split())\n",
    "\n",
    "    model = gensim.models.Word2Vec(sentences, min_count=10)\n",
    "    model = gensim.models.Word2Vec(sentences, size=300,workers=core_count)\n",
    "\n",
    "    model.save('../DataFiles/trained_vec.bin')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Loading the model, \n",
    "## word similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2017-01-24 14:40:22,265 : INFO : loading Word2Vec object from ../DataFiles/trained_vec.bin\n",
      "2017-01-24 14:40:22,669 : INFO : loading syn1neg from ../DataFiles/trained_vec.bin.syn1neg.npy with mmap=None\n",
      "2017-01-24 14:40:23,532 : INFO : loading syn0 from ../DataFiles/trained_vec.bin.syn0.npy with mmap=None\n",
      "2017-01-24 14:40:24,459 : INFO : setting ignored attribute syn0norm to None\n",
      "2017-01-24 14:40:24,460 : INFO : setting ignored attribute cum_table to None\n",
      "2017-01-24 14:40:24,461 : INFO : loaded ../DataFiles/trained_vec.bin\n",
      "2017-01-24 14:40:24,625 : INFO : precomputing L2-norms of word weight vectors\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>word</th>\n",
       "      <th>similarity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>cottage</td>\n",
       "      <td>0.802360</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>bungalow</td>\n",
       "      <td>0.751224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>guesthouse</td>\n",
       "      <td>0.747456</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>casita</td>\n",
       "      <td>0.712503</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>camper</td>\n",
       "      <td>0.686172</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>airstream</td>\n",
       "      <td>0.685671</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>studio</td>\n",
       "      <td>0.673551</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>hideaway</td>\n",
       "      <td>0.663801</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>trailer</td>\n",
       "      <td>0.660531</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>treehouse</td>\n",
       "      <td>0.657508</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         word  similarity\n",
       "0     cottage    0.802360\n",
       "1    bungalow    0.751224\n",
       "2  guesthouse    0.747456\n",
       "3      casita    0.712503\n",
       "4      camper    0.686172\n",
       "5   airstream    0.685671\n",
       "6      studio    0.673551\n",
       "7    hideaway    0.663801\n",
       "8     trailer    0.660531\n",
       "9   treehouse    0.657508"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import gensim\n",
    "model = gensim.models.Word2Vec.load('../DataFiles/trained_vec.bin')\n",
    "\n",
    "df = pd.DataFrame(model.similar_by_word('cabin',topn=10))\n",
    "df.columns=[['word','similarity']]\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "## Algebra equation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('hostess', 0.804054856300354)]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar(positive=['host', 'woman'], negative=['man'], topn=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Matching"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'welcome'"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.doesnt_match(\"wall towel bathroom welcome cabin\".split())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Similartity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.267094335798\n",
      "0.633313943257\n"
     ]
    }
   ],
   "source": [
    "print(model.similarity('cabin', 'inn'))\n",
    "print(model.similarity('hostel', 'hotel'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Ploting word vectors in 2D using PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.28550961  0.220871  ]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhEAAAIDCAYAAAC6gEyCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x114a21110>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "words = [\"block\",\"station\",\"train\",\"subway\",\"dog\",\"cat\",\"restaurant\",\"eat\",'lunch','breakfast']\n",
    "vectors=[]\n",
    "for word in words:\n",
    "    vectors.append(model[word])\n",
    "sns.set_style(\"darkgrid\")\n",
    "plot1(words,vectors)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Raw vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 1.4550308 ,  0.3940624 ,  0.96272272, -0.51513439,  0.05838453,\n",
       "        0.11652667, -1.28175569, -0.84822196,  0.26959687,  0.3268182 ,\n",
       "        0.21388273, -0.29465479, -0.88845855,  2.33686543, -0.36153826,\n",
       "        1.60229075,  0.14986882,  0.72073853,  0.04800477,  1.66996908,\n",
       "       -0.17910798,  0.87562811,  1.31664038,  0.49429235,  0.91698527,\n",
       "        0.04697718,  0.22947681,  0.90869707,  0.72565889, -0.12199731,\n",
       "        0.01736214, -1.58636618,  1.12740386,  0.62459791, -0.03183404,\n",
       "       -0.10394789, -0.6875841 , -1.09427857,  2.89352727, -0.13843235,\n",
       "        1.65345323,  1.71826708, -1.10500216,  0.30762517,  0.13389966,\n",
       "       -0.4965795 ,  2.86934876,  1.85880089, -0.80497146, -0.23941553,\n",
       "        0.58636546, -1.78203833, -0.55848557,  0.52863246,  0.77804691,\n",
       "        0.31252599,  0.26360765, -1.10693419,  0.54123145, -0.20988232,\n",
       "        0.69677615,  0.72582155,  1.05308759, -0.1398433 , -0.14856255,\n",
       "       -2.5158124 ,  1.71526957, -1.04685533, -0.55286878, -0.1129133 ,\n",
       "        2.01186848,  0.54050422,  1.31919563, -0.38752496,  1.37044621,\n",
       "       -0.58384275, -1.72821212,  0.65290874,  1.02914107, -1.75459445,\n",
       "       -0.65472674, -0.51760757, -0.275406  , -1.1399188 ,  0.88444769,\n",
       "       -0.54059672, -0.32805946,  0.87617445,  0.17066278, -0.3035174 ,\n",
       "        2.27022719, -1.73183024, -0.75178719,  2.29013324, -0.94638109,\n",
       "       -0.13546449, -0.04058611,  0.60500282,  0.08734755,  0.81744605,\n",
       "        1.23542917,  0.79620314,  0.22806773,  0.50903082,  0.86773896,\n",
       "        1.30616283, -0.36672011,  1.73984981, -0.52320522,  0.39040953,\n",
       "       -0.88937181, -0.12129363,  1.50743294, -1.18421483, -2.16254783,\n",
       "        0.87187839,  2.24252582, -0.59691799, -0.64582247,  0.05996878,\n",
       "       -0.35840046,  0.60517031, -1.29956317, -0.05574429, -0.40261576,\n",
       "        1.69508016, -0.64350641, -0.13833967, -0.13993761,  0.76335025,\n",
       "       -0.98122406, -0.78364182,  0.15019812,  0.21855225,  0.95313734,\n",
       "        0.5360254 ,  0.6490193 , -0.13839862,  0.35301158,  0.66604918,\n",
       "       -0.13319796, -1.77688241,  0.69028109, -0.7523573 , -0.77802449,\n",
       "        2.01728678, -1.57277024,  0.12080555,  1.74674737, -0.13986866,\n",
       "        1.90172148,  1.34654689, -0.20901887,  1.66882014,  0.23160565,\n",
       "       -1.82262802, -1.65439379, -1.20238292,  0.4243333 ,  0.48295772,\n",
       "        0.31050205, -1.39215767,  1.72040069, -1.43318009, -0.42201972,\n",
       "        0.74081784, -0.50901854,  2.3682816 , -1.72443914,  1.6914885 ,\n",
       "       -1.61578751, -0.76086122,  2.81549025, -2.18755126, -2.47492957,\n",
       "       -2.91821432,  0.27839521, -0.19366249,  0.46679592, -1.51271951,\n",
       "        1.10492885,  0.18609767, -1.62350309, -0.44507897, -0.36114183,\n",
       "        0.43592709, -1.19844103, -0.40497223, -1.42174983, -0.4582389 ,\n",
       "        0.07030702,  0.55136234, -0.60940725,  0.33863997,  0.81235266,\n",
       "        0.49644291,  0.06793395,  0.75775057,  0.52719069, -0.61217052,\n",
       "        1.80208063,  1.62443399, -1.67111766, -2.35280037, -1.25490391,\n",
       "        0.46405375,  2.43335485,  3.0033946 , -0.46108857, -1.52367759,\n",
       "        1.93958998,  1.44639552, -1.76742625, -1.42761528,  1.12414753,\n",
       "        0.64531654, -1.94550312, -2.58154321,  0.81433934, -0.88705069,\n",
       "        0.57639873, -1.65421844,  0.80111688, -0.16415867, -0.29469606,\n",
       "        0.02186927,  0.01760439, -1.30025208,  2.32926059, -2.17580581,\n",
       "        0.93974787,  0.08391468,  1.09198213, -0.25365904, -0.8804552 ,\n",
       "        1.28282654, -0.41716719, -0.0294174 ,  0.49726272,  1.20480871,\n",
       "        1.55795825, -0.00915126,  1.11130512, -0.08960282, -0.81888402,\n",
       "        0.28353992, -0.29784259,  1.1847471 , -0.0255003 ,  0.07153933,\n",
       "        0.69996476,  1.47916484, -0.20856088,  1.72833073,  1.28614819,\n",
       "       -0.13222463, -0.15864006, -0.65887648,  1.24914479,  0.5725112 ,\n",
       "       -1.64166164,  0.5549131 , -0.33932999,  0.37700921,  1.72460568,\n",
       "       -0.43587673, -1.11459661,  0.22944088,  0.34784216,  0.1645131 ,\n",
       "        0.7203601 , -0.12187007,  1.73521304,  2.20218921, -0.73274535,\n",
       "        0.32963142, -0.98893064, -1.68289852, -3.0325067 ,  0.18043333,\n",
       "       -1.69703507, -1.41218388, -0.68998265, -1.79191935, -0.62326252,\n",
       "       -0.94067103, -1.39435399,  2.5108428 ,  2.92108893,  1.46405745,\n",
       "       -0.12491181,  0.51868832,  0.91126662,  0.1823824 , -0.06128192,\n",
       "        0.38681501,  0.51733977, -0.08270054,  0.4003472 , -1.33066881], dtype=float32)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model['guest'] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}