{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true,
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [
    {
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1-1cc026d5c8bb>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mnumpy\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mtensorflow\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0mcorpus_raw\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m'He is the king . The king is royal . She is the royal  queen '\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;31m# convert to lower case\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mcorpus_raw\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcorpus_raw\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlower\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'tensorflow'"
     ],
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'tensorflow'",
     "output_type": "error"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "corpus_raw = 'He is the king . The king is royal . She is the royal  queen '\n",
    "# convert to lower case\n",
    "corpus_raw = corpus_raw.lower()\n",
    "#print(corpus_raw)\n",
    "words = []\n",
    "for word in corpus_raw.split():\n",
    "    if word != '.': # because we don't want to treat . as a word\n",
    "        words.append(word)\n",
    "words = set(words) # so that all duplicate words are removed\n",
    "word2int = {}\n",
    "int2word = {}\n",
    "vocab_size = len(words) # gives the total number of unique words\n",
    "for i,word in enumerate(words):\n",
    "    word2int[word] = i\n",
    "    int2word[i] = word\n",
    "#print(word2int['queen'])\n",
    "#print(int2word[5])\n",
    "\n",
    "# raw sentences is a list of sentences.\n",
    "raw_sentences = corpus_raw.split('.')\n",
    "sentences = []\n",
    "for sentence in raw_sentences:\n",
    "    sentences.append(sentence.split())\n",
    "#print(sentences)\n",
    "data = []\n",
    "WINDOW_SIZE = 2\n",
    "for sentence in sentences:\n",
    "    for word_index, word in enumerate(sentence):\n",
    "        for nb_word in sentence[max(word_index - WINDOW_SIZE, 0) : min(word_index + WINDOW_SIZE, len(sentence)) + 1] : \n",
    "            if nb_word != word:\n",
    "                data.append([word, nb_word])\n",
    "#print(data)\n",
    "\n",
    "# function to convert numbers to one hot vectors\n",
    "def to_one_hot(data_point_index, vocab_size):\n",
    "    temp = np.zeros(vocab_size)\n",
    "    temp[data_point_index] = 1\n",
    "    return temp\n",
    "x_train = [] # input word\n",
    "y_train = [] # output word\n",
    "for data_word in data:\n",
    "    x_train.append(to_one_hot(word2int[ data_word[0] ], vocab_size))\n",
    "    y_train.append(to_one_hot(word2int[ data_word[1] ], vocab_size))\n",
    "# convert them to numpy arrays\n",
    "x_train = np.asarray(x_train)\n",
    "y_train = np.asarray(y_train)\n",
    "\n",
    "print(x_train.shape, y_train.shape)\n",
    "x = tf.placeholder(tf.float32, shape=(None, vocab_size))\n",
    "y_label = tf.placeholder(tf.float32, shape=(None, vocab_size))\n",
    "\n",
    "EMBEDDING_DIM = 5 # you can choose your own number\n",
    "W1 = tf.Variable(tf.random_normal([vocab_size, EMBEDDING_DIM]))\n",
    "b1 = tf.Variable(tf.random_normal([EMBEDDING_DIM])) #bias\n",
    "hidden_representation = tf.add(tf.matmul(x,W1), b1)\n",
    "W2 = tf.Variable(tf.random_normal([EMBEDDING_DIM, vocab_size]))\n",
    "b2 = tf.Variable(tf.random_normal([vocab_size]))\n",
    "prediction = tf.nn.softmax(tf.add( tf.matmul(hidden_representation, W2), b2))\n",
    "\n",
    "sess = tf.Session()\n",
    "init = tf.global_variables_initializer()\n",
    "sess.run(init) #make sure you do this!\n",
    "# define the loss function:\n",
    "cross_entropy_loss = tf.reduce_mean(-tf.reduce_sum(y_label * tf.log(prediction), reduction_indices=[1]))\n",
    "# define the training step:\n",
    "train_step = tf.train.GradientDescentOptimizer(0.1).minimize(cross_entropy_loss)\n",
    "n_iters = 10000\n",
    "# train for n_iter iterations\n",
    "for _ in range(n_iters):\n",
    "    sess.run(train_step, feed_dict={x: x_train, y_label: y_train})\n",
    "    #print('loss is : ', sess.run(cross_entropy_loss, feed_dict={x: x_train, y_label: y_train}))\n",
    "#print(sess.run(W1))\n",
    "#print('----------')\n",
    "#print(sess.run(b1))\n",
    "#print('----------')\n",
    "vectors = sess.run(W1 + b1)\n",
    "#print(vectors)\n",
    "print(vectors[ word2int['queen'] ])"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  },
  "kernelspec": {
   "name": "python3",
   "language": "python",
   "display_name": "Python 3"
  },
  "pycharm": {
   "stem_cell": {
    "cell_type": "raw",
    "source": [],
    "metadata": {
     "collapsed": false
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}