vocabulary-embedding.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generate intial word embedding for headlines and description"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The embedding is limited to a fixed vocabulary size (`vocab_size`) but\n",
    "a vocabulary of all the words that appeared in the data is built."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "FN = 'vocabulary-embedding'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "seed=42"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "vocab_size = 40000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "embedding_dim = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "lower = False # dont lower case the text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# read tokenized headlines and descriptions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import cPickle as pickle\n",
    "FN0 = 'tokens' # this is the name of the data file which I assume you already have\n",
    "with open('data/%s.pkl'%FN0, 'rb') as fp:\n",
    "    heads, desc, keywords = pickle.load(fp) # keywords are not used in this project"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "if lower:\n",
    "    heads = [h.lower() for h in heads]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "if lower:\n",
    "    desc = [h.lower() for h in desc]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Remainders : Super wi-fi edition'"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "i=0\n",
    "heads[i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"! Google 's free San Fran wi-fi will track your location . Oh , do n't worry , they have some clause about evil , right ? ! Silicon Valley , meanwhile , plans its own wi-fi coverage . It 'll be like San Fran 's , only with fewer homeless trustafarians using it . ! Soon you 'll have wi-fi on the BART . Caveat : After a few days on those seats , your laptop will smell like urine . ! Boy in the Bubble : The sad story\""
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "desc[i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Google',\n",
       " 'Gawker',\n",
       " 'Remainders',\n",
       " 'WiFi',\n",
       " 'Valleywag',\n",
       " 'Nyt',\n",
       " 'Vcs',\n",
       " 'San Francisco']"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keywords[i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(684114, 672187)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(heads),len(set(heads))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(684114, 683908)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(desc),len(set(desc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# build vocabulary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from collections import Counter\n",
    "from itertools import chain\n",
    "def get_vocab(lst):\n",
    "    vocabcount = Counter(w for txt in lst for w in txt.split())\n",
    "    vocab = map(lambda x: x[0], sorted(vocabcount.items(), key=lambda x: -x[1]))\n",
    "    return vocab, vocabcount"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "vocab, vocabcount = get_vocab(heads+desc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "most popular tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[',', '.', 'the', 'a', 'to', 'of', \"'s\", 'and', 'in', 'The', 'is', 'for', 'that', ':', 'on', \"''\", '``', 'you', 'it', 'with', '?', 'at', \"n't\", 'this', \"'\", '(', 'was', ')', 'has', 'be', 'are', 'from', 'as', '--', 'A', 'an', 'I', 'his', '!', 'have', 'by', 'but', 'It', 'In', 'we', 'your', 'To', 'about', 'he', 'You']\n",
      "... 523732\n"
     ]
    }
   ],
   "source": [
    "print vocab[:50]\n",
    "print '...',len(vocab)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/udi/anaconda/lib/python2.7/site-packages/matplotlib/__init__.py:872: UserWarning: axes.color_cycle is deprecated and replaced with axes.prop_cycle; please use the latter.\n",
      "  warnings.warn(self.msg_depr % (key, alt_key))\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAEhCAYAAABRKfYcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGsQqEGOMMQ1iFYgxxpgGsQrEGGNMg/w/4MrDSGMnEIgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10ef52f90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.plot([vocabcount[w] for w in vocab]);\n",
    "plt.gca().set_xscale(\"log\", nonposx='clip')\n",
    "plt.gca().set_yscale(\"log\", nonposy='clip')\n",
    "plt.title('word distribution in headlines and discription')\n",
    "plt.xlabel('rank')\n",
    "plt.ylabel('total appearances');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "always nice to see [Zipf's law](https://en.wikipedia.org/wiki/Zipf%27s_law)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Index words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "empty = 0 # RNN mask of no data\n",
    "eos = 1  # end of sentence\n",
    "start_idx = eos+1 # first real word"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_idx(vocab, vocabcount):\n",
    "    word2idx = dict((word, idx+start_idx) for idx,word in enumerate(vocab))\n",
    "    word2idx['<empty>'] = empty\n",
    "    word2idx['<eos>'] = eos\n",
    "    \n",
    "    idx2word = dict((idx,word) for word,idx in word2idx.iteritems())\n",
    "\n",
    "    return word2idx, idx2word"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "word2idx, idx2word = get_idx(vocab, vocabcount)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Word Embedding"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## read GloVe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fname = 'glove.6B.%dd.txt'%embedding_dim\n",
    "import os\n",
    "datadir_base = os.path.expanduser(os.path.join('~', '.keras'))\n",
    "if not os.access(datadir_base, os.W_OK):\n",
    "    datadir_base = os.path.join('/tmp', '.keras')\n",
    "datadir = os.path.join(datadir_base, 'datasets')\n",
    "glove_name = os.path.join(datadir, fname)\n",
    "if not os.path.exists(glove_name):\n",
    "    path = 'glove.6B.zip'\n",
    "    path = get_file(path, origin=\"http://nlp.stanford.edu/data/glove.6B.zip\")\n",
    "    !unzip {datadir}/{path}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "400000"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "glove_n_symbols = !wc -l {glove_name}\n",
    "glove_n_symbols = int(glove_n_symbols[0].split()[0])\n",
    "glove_n_symbols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "glove_index_dict = {}\n",
    "glove_embedding_weights = np.empty((glove_n_symbols, embedding_dim))\n",
    "globale_scale=.1\n",
    "with open(glove_name, 'r') as fp:\n",
    "    i = 0\n",
    "    for l in fp:\n",
    "        l = l.strip().split()\n",
    "        w = l[0]\n",
    "        glove_index_dict[w] = i\n",
    "        glove_embedding_weights[i,:] = map(float,l[1:])\n",
    "        i += 1\n",
    "glove_embedding_weights *= globale_scale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.040815727600190289"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "glove_embedding_weights.std()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "for w,i in glove_index_dict.iteritems():\n",
    "    w = w.lower()\n",
    "    if w not in glove_index_dict:\n",
    "        glove_index_dict[w] = i"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## embedding matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "use GloVe to initialize embedding matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "random-embedding/glove scale 0.0706949139514 std 0.0408138249575\n",
      "number of tokens, in small vocab, found in glove and copied to embedding 39332 0.9833\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "# generate random embedding with same scale as glove\n",
    "np.random.seed(seed)\n",
    "shape = (vocab_size, embedding_dim)\n",
    "scale = glove_embedding_weights.std()*np.sqrt(12)/2 # uniform and not normal\n",
    "embedding = np.random.uniform(low=-scale, high=scale, size=shape)\n",
    "print 'random-embedding/glove scale', scale, 'std', embedding.std()\n",
    "\n",
    "# copy from glove weights of words that appear in our short vocabulary (idx2word)\n",
    "c = 0\n",
    "for i in range(vocab_size):\n",
    "    w = idx2word[i]\n",
    "    g = glove_index_dict.get(w, glove_index_dict.get(w.lower()))\n",
    "    if g is None and w.startswith('#'): # glove has no hastags (I think...)\n",
    "        w = w[1:]\n",
    "        g = glove_index_dict.get(w, glove_index_dict.get(w.lower()))\n",
    "    if g is not None:\n",
    "        embedding[i,:] = glove_embedding_weights[g,:]\n",
    "        c+=1\n",
    "print 'number of tokens, in small vocab, found in glove and copied to embedding', c,c/float(vocab_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "lots of word in the full vocabulary (word2idx) are outside `vocab_size`.\n",
    "Build an alterantive which will map them to their closest match in glove but only if the match\n",
    "is good enough (cos distance above `glove_thr`)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "glove_thr = 0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "word2glove = {}\n",
    "for w in word2idx:\n",
    "    if w in glove_index_dict:\n",
    "        g = w\n",
    "    elif w.lower() in glove_index_dict:\n",
    "        g = w.lower()\n",
    "    elif w.startswith('#') and w[1:] in glove_index_dict:\n",
    "        g = w[1:]\n",
    "    elif w.startswith('#') and w[1:].lower() in glove_index_dict:\n",
    "        g = w[1:].lower()\n",
    "    else:\n",
    "        continue\n",
    "    word2glove[w] = g"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "for every word outside the embedding matrix find the closest word inside the mebedding matrix.\n",
    "Use cos distance of GloVe vectors.\n",
    "\n",
    "Allow for the last `nb_unknown_words` words inside the embedding matrix to be considered to be outside.\n",
    "Dont accept distances below `glove_thr`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of glove substitutes found 122377\n"
     ]
    }
   ],
   "source": [
    "normed_embedding = embedding/np.array([np.sqrt(np.dot(gweight,gweight)) for gweight in embedding])[:,None]\n",
    "\n",
    "nb_unknown_words = 100\n",
    "\n",
    "glove_match = []\n",
    "for w,idx in word2idx.iteritems():\n",
    "    if idx >= vocab_size-nb_unknown_words and w.isalpha() and w in word2glove:\n",
    "        gidx = glove_index_dict[word2glove[w]]\n",
    "        gweight = glove_embedding_weights[gidx,:].copy()\n",
    "        # find row in embedding that has the highest cos score with gweight\n",
    "        gweight /= np.sqrt(np.dot(gweight,gweight))\n",
    "        score = np.dot(normed_embedding[:vocab_size-nb_unknown_words], gweight)\n",
    "        while True:\n",
    "            embedding_idx = score.argmax()\n",
    "            s = score[embedding_idx]\n",
    "            if s < glove_thr:\n",
    "                break\n",
    "            if idx2word[embedding_idx] in word2glove :\n",
    "                glove_match.append((w, embedding_idx, s)) \n",
    "                break\n",
    "            score[embedding_idx] = -1\n",
    "glove_match.sort(key = lambda x: -x[2])\n",
    "print '# of glove substitutes found', len(glove_match)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "manually check that the worst substitutions we are going to do are good enough"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.500009203566 CQ => Deletes\n",
      "0.500009203566 cq => Deletes\n",
      "0.500005559501 cheapskate => dude\n",
      "0.500005559501 Cheapskate => dude\n",
      "0.500004635765 Stoney => Ridgeway\n",
      "0.500004635765 stoney => Ridgeway\n",
      "0.500003799 Maia => Joanna\n",
      "0.500003553879 Jammy => fruity\n",
      "0.500003553879 jammy => fruity\n",
      "0.500003378034 ataxia => OCD\n"
     ]
    }
   ],
   "source": [
    "for orig, sub, score in glove_match[-10:]:\n",
    "    print score, orig,'=>', idx2word[sub]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "build a lookup table of index of outside words to index of inside words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "glove_idx2idx = dict((word2idx[w],embedding_idx) for  w, embedding_idx, _ in glove_match)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "684114"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y = [[word2idx[token] for token in headline.split()] for headline in heads]\n",
    "len(Y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYcAAAEACAYAAABYq7oeAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x137aad710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(map(len,Y),bins=50);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "684114"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = [[word2idx[token] for token in d.split()] for d in desc]\n",
    "len(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEACAYAAACtVTGuAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x110751d50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(map(len,X),bins=50);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import cPickle as pickle\n",
    "with open('data/%s.pkl'%FN,'wb') as fp:\n",
    "    pickle.dump((embedding, idx2word, word2idx, glove_idx2idx),fp,-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import cPickle as pickle\n",
    "with open('data/%s.data.pkl'%FN,'wb') as fp:\n",
    "    pickle.dump((X,Y),fp,-1)"
   ]
  }
 ],
 "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": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}