nlp_word2vec.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Gensim word embedding"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Word embedding is the collective name for a set of language modeling and feature learning techniques in natural language processing (NLP) where words or phrases from the vocabulary are mapped to vectors of real numbers. Conceptually it involves a mathematical embedding from a space with one dimension per word to a continuous vector space with much lower dimension.\n",
    "\n",
    "Word2vec takes as its input a large corpus of text and produces a vector space, typically of several hundred dimensions, with each unique word in the corpus being assigned a corresponding vector in the space.\n",
    "\n",
    "Word vectors are positioned in the vector space such that words that share common contexts in the corpus are located in close proximity to one another in the space. The purpose and usefulness of Word2vec is to group the vectors of similar words together in vectorspace.\n",
    "\n",
    "The points have some nice properties (words with similar meanings tend to occur closer to each other) [proximity is measured using cosine distance between 2 word vectors]\n",
    "\n",
    "You can use on of the performance evaluation methods, such as the Word error rate (WER) or Preplexity (PPR).\n",
    "\n",
    "Word2vec training is an unsupervised task, there’s no good way to objectively evaluate the result. Evaluation depends on your end application."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "from matplotlib import pyplot as plt\n",
    "import pandas as pd\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define training data\n",
    "sentences = [\n",
    "            ['this', 'is', 'the', 'first', 'sentence', 'for', 'word2vec'],\n",
    "            ['this', 'is', 'the', 'second', 'sentence'],\n",
    "            ['yet', 'another', 'sentence'],\n",
    "            ['one', 'more', 'sentence'],\n",
    "            ['and', 'the', 'final', 'sentence']\n",
    "            ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Gensim word2vec model\n",
    "\n",
    "# min_count: Words below the min_count frequency are dropped before training occurs\n",
    "\n",
    "# size : Dimensionality of the feature vectors (normal range 100 to 300)\n",
    "\n",
    "# sg : skip-gram or CBOW\n",
    "\n",
    "from gensim.models import Word2Vec\n",
    "model = Word2Vec(sentences, size=300, min_count=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'this': <gensim.models.keyedvectors.Vocab at 0x49dd8e0>,\n",
       " 'is': <gensim.models.keyedvectors.Vocab at 0x49ddd60>,\n",
       " 'the': <gensim.models.keyedvectors.Vocab at 0x49dd808>,\n",
       " 'first': <gensim.models.keyedvectors.Vocab at 0x49dde98>,\n",
       " 'sentence': <gensim.models.keyedvectors.Vocab at 0x49dd658>,\n",
       " 'for': <gensim.models.keyedvectors.Vocab at 0x49dde50>,\n",
       " 'word2vec': <gensim.models.keyedvectors.Vocab at 0x49dd490>,\n",
       " 'second': <gensim.models.keyedvectors.Vocab at 0x49dd778>,\n",
       " 'yet': <gensim.models.keyedvectors.Vocab at 0x49ddef8>,\n",
       " 'another': <gensim.models.keyedvectors.Vocab at 0x49ddd00>,\n",
       " 'one': <gensim.models.keyedvectors.Vocab at 0x49dddd8>,\n",
       " 'more': <gensim.models.keyedvectors.Vocab at 0x49ddb38>,\n",
       " 'and': <gensim.models.keyedvectors.Vocab at 0x49dd898>,\n",
       " 'final': <gensim.models.keyedvectors.Vocab at 0x49ddd30>}"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.wv.vocab # wv objects contain mapping b/w words and embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 9.3474059e-04, -1.2397201e-03, -5.7488465e-04, ...,\n",
       "         4.7587953e-04, -1.7098231e-04,  5.1608131e-05],\n",
       "       [ 5.2183220e-04, -2.8954138e-04,  1.5949433e-03, ...,\n",
       "        -1.4195421e-04,  2.0076369e-04, -6.2771322e-04],\n",
       "       [ 9.4563048e-04,  7.4043270e-04, -1.2772779e-03, ...,\n",
       "         1.0228147e-03,  1.4452571e-03, -1.0443116e-03],\n",
       "       ...,\n",
       "       [-7.1219128e-04,  5.2793475e-04, -5.9428083e-04, ...,\n",
       "         7.7968405e-04,  6.8202015e-04,  5.2004913e-04],\n",
       "       [ 1.9807000e-04,  1.3697292e-03,  1.2804329e-03, ...,\n",
       "         1.1800590e-03, -5.8450934e-04, -3.9673451e-04],\n",
       "       [-9.4368780e-04,  1.3291637e-03, -5.1989371e-04, ...,\n",
       "         1.1719497e-03,  3.4152597e-04,  1.3979293e-03]], dtype=float32)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model[model.wv.vocab]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('word2vec', 0.07985182851552963),\n",
       " ('sentence', 0.036238837987184525),\n",
       " ('more', 0.031186960637569427),\n",
       " ('another', 0.030362768098711967),\n",
       " ('this', -0.0013018168974667788),\n",
       " ('one', -0.00695493258535862),\n",
       " ('final', -0.010404408909380436),\n",
       " ('yet', -0.019741125404834747),\n",
       " ('and', -0.03142162039875984),\n",
       " ('second', -0.03764071315526962)]"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.wv.most_similar(\"first\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-4.80971066e-04,  6.85635372e-04, -1.45494658e-03, -1.20409706e-03,\n",
       "       -2.66982475e-04,  1.36695779e-03, -2.25089505e-04, -5.67820098e-04,\n",
       "        3.70433001e-04, -1.48384995e-03,  6.08932642e-05,  2.23350304e-04,\n",
       "       -6.63744227e-04,  1.35649019e-03,  1.11396264e-04,  3.01766180e-04,\n",
       "       -1.48460816e-03,  1.26197422e-03, -1.65242050e-03, -1.27496116e-03,\n",
       "       -1.73614026e-04,  7.60554758e-05,  1.30074588e-03,  7.00145960e-04,\n",
       "       -1.38576771e-03, -1.45148297e-04, -1.38108619e-03,  1.01223739e-03,\n",
       "       -1.08501315e-03, -1.24420412e-03,  3.69852089e-04,  5.02401308e-05,\n",
       "       -7.42561242e-04,  1.53951661e-03, -7.20208336e-04, -1.07680995e-03,\n",
       "        6.04422938e-04,  2.56982021e-04,  1.18024764e-04, -1.56695372e-04,\n",
       "        3.61078361e-04, -1.88029357e-04, -3.50765506e-04, -1.05043838e-03,\n",
       "        9.77800926e-04,  1.10251363e-03,  5.31006604e-04,  1.56616524e-03,\n",
       "       -1.34955929e-03,  8.78855644e-05,  5.57851687e-04, -5.59356762e-04,\n",
       "       -1.08118134e-03,  1.53029279e-03,  4.15198592e-04,  3.95941257e-04,\n",
       "        1.57634204e-03,  1.65542209e-04,  3.39456979e-04, -1.50352344e-03,\n",
       "       -1.14361825e-03,  1.54452166e-03,  1.57936011e-05, -9.59791301e-04,\n",
       "       -6.18393125e-04,  2.40288646e-04, -4.49977291e-04, -1.59583439e-03,\n",
       "        3.60984355e-04, -1.46811747e-03, -8.36117193e-04,  7.59725444e-05,\n",
       "        1.10883208e-03, -6.09726063e-04,  3.45373293e-04, -5.89658506e-04,\n",
       "        9.24355641e-04, -1.58353173e-03,  1.23229879e-03,  1.07455009e-03,\n",
       "        3.40717554e-04, -4.30048443e-04,  1.50843838e-03,  9.06191010e-04,\n",
       "       -1.23620685e-03,  7.36508460e-04, -5.52524230e-04, -1.23354432e-03,\n",
       "        8.70609001e-05, -1.13072863e-03,  1.45506265e-03, -1.52418821e-03,\n",
       "        2.18590241e-04, -1.27342495e-03, -3.43198801e-04, -6.78795215e-04,\n",
       "       -1.46942923e-03, -1.29731349e-03,  7.20606127e-04, -5.70822740e-04,\n",
       "        1.43330824e-03, -1.23730593e-03, -9.76612209e-04,  8.74104095e-04,\n",
       "       -1.88364575e-04,  1.66849597e-04, -1.23603141e-03,  1.61335967e-03,\n",
       "        7.25183112e-04, -1.65144203e-03, -3.70543275e-04, -1.33965048e-03,\n",
       "        5.18125540e-04,  6.56132179e-04, -2.54992774e-05,  1.02369220e-03,\n",
       "        1.22282095e-03,  6.88234170e-04,  5.05388016e-04, -1.37261976e-03,\n",
       "       -7.33794470e-04,  7.59374932e-04, -3.33758275e-04, -1.21101388e-03,\n",
       "        9.42231505e-04, -1.44734466e-03,  1.61894865e-03,  9.20801889e-04,\n",
       "        1.13958272e-03, -1.38235837e-03, -4.81461291e-04, -1.62103307e-03,\n",
       "       -1.19974822e-04,  1.44425558e-03,  5.57553431e-04, -1.61719450e-03,\n",
       "        4.41902142e-04,  1.41608936e-03, -1.55346817e-03, -4.44608362e-04,\n",
       "       -1.34220079e-03,  7.62921467e-04,  7.88524048e-04, -6.55125885e-04,\n",
       "        1.22484425e-03, -8.64029687e-04,  1.13017777e-04, -8.53469130e-04,\n",
       "        4.84301345e-05,  5.08958416e-04,  1.16411806e-03,  8.78458435e-04,\n",
       "        5.78804233e-04, -1.04642659e-03,  1.10630039e-03,  7.71489926e-04,\n",
       "        1.50014134e-03, -7.40472518e-04,  9.28994195e-05, -5.88385388e-04,\n",
       "        8.39841960e-04,  6.19689527e-04, -1.27960346e-04,  1.63093524e-03,\n",
       "       -9.00384388e-04, -2.63520014e-05, -4.33687877e-04,  2.21363865e-04,\n",
       "        1.14339031e-03, -1.52172986e-03,  8.57139123e-04, -7.67615449e-04,\n",
       "        9.33882606e-04,  1.03774120e-03, -3.62668885e-04,  7.13259389e-04,\n",
       "        6.87213091e-04,  8.03361356e-04, -5.79029846e-04,  4.55925125e-04,\n",
       "       -1.46169288e-04,  9.68954817e-04,  1.16487674e-03,  2.22405477e-04,\n",
       "        4.75068053e-04, -5.18133864e-04, -1.22063945e-03, -1.20944262e-03,\n",
       "        1.45671470e-03, -4.91999788e-04,  3.34027136e-04, -2.70281773e-04,\n",
       "        1.08042290e-03, -7.05427374e-04,  6.15197117e-04, -3.00941116e-04,\n",
       "       -3.84248473e-04, -1.20603514e-03,  7.52302993e-04, -6.34672644e-04,\n",
       "        1.49165501e-03, -1.25384412e-03, -7.85767857e-04, -1.01134181e-03,\n",
       "       -8.12749844e-04, -2.69263110e-04,  2.14669752e-04,  9.39672536e-05,\n",
       "       -1.38962641e-04, -1.09714536e-04, -1.16763182e-03,  3.02851608e-04,\n",
       "        5.86256152e-04,  3.74441908e-04, -1.45962404e-03, -9.84697370e-04,\n",
       "       -6.03638997e-04,  8.19628884e-04,  4.24288126e-04,  1.41731091e-03,\n",
       "        6.19657367e-05,  1.42274424e-03,  1.43538869e-03, -1.24108722e-03,\n",
       "       -1.89707323e-04, -4.24811878e-04,  7.26782542e-04, -1.40004279e-03,\n",
       "        1.30510191e-03, -2.83189816e-04, -1.55855203e-03, -7.21776625e-04,\n",
       "       -1.44640985e-03,  1.37027132e-03,  1.57666020e-03, -1.44814572e-03,\n",
       "       -2.86909722e-04,  1.66419777e-03, -1.47226593e-03, -1.43955974e-03,\n",
       "       -4.10855864e-04, -1.28818396e-03, -5.87622017e-05, -8.82572902e-04,\n",
       "       -1.63566670e-03, -1.61803604e-04, -7.53150613e-04, -7.58583250e-04,\n",
       "        1.17386191e-03, -5.61388384e-04,  1.55002042e-03,  8.67832568e-04,\n",
       "        1.51024899e-03, -1.22267765e-03,  7.83226802e-04, -3.21421394e-04,\n",
       "        1.35223242e-03,  1.41103484e-03,  1.12625468e-03, -8.52788973e-04,\n",
       "       -1.41908834e-03, -6.38248457e-04, -1.59390189e-03,  3.66584427e-04,\n",
       "        3.62396735e-04, -3.43733176e-04, -1.31213153e-03, -2.65053852e-04,\n",
       "       -3.12494114e-04, -2.55864761e-05, -1.20058702e-03, -6.34376251e-04,\n",
       "        9.83409700e-04,  4.77290814e-05, -1.62198034e-04,  1.13898959e-05,\n",
       "        1.02465344e-03,  4.53879795e-04, -1.44339690e-03, -8.16293643e-04,\n",
       "       -9.96720279e-04, -8.79695930e-04,  1.25831910e-04,  1.63981970e-03,\n",
       "        9.69633809e-04,  1.55796914e-03,  2.20043148e-04, -7.80179515e-04,\n",
       "        3.96241812e-04,  1.77912007e-04,  1.47097604e-03, -1.31815963e-03,\n",
       "       -1.44679251e-03,  7.13929767e-04, -2.40739828e-04, -1.43821572e-03,\n",
       "       -9.29246948e-04,  2.68115400e-04, -5.68661373e-04, -3.11042385e-07],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.wv['first'] # dimensionality or size of feature vecotor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualising the Word Embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "# fit a 2d PCA model to the vectors\n",
    "X = model[model.wv.vocab]\n",
    "pca = PCA(n_components=2)\n",
    "result = pca.fit_transform(X)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PCA visualization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "# create a scatter plot of the projection\n",
    "plt.scatter(result[:, 0], result[:, 1])\n",
    "words = list(model.wv.vocab)\n",
    "\n",
    "for i, word in enumerate(words):\n",
    "    plt.annotate(word, xy=(result[i, 0], result[i, 1]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## t-SNE Visualization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "tsne = TSNE(n_components=2).fit_transform(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>TSNE component 1</th>\n",
       "      <th>TSNE component 2</th>\n",
       "      <th>word</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>266.298431</td>\n",
       "      <td>23.642027</td>\n",
       "      <td>this</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>132.896149</td>\n",
       "      <td>-161.362015</td>\n",
       "      <td>is</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-329.322601</td>\n",
       "      <td>-13.235461</td>\n",
       "      <td>the</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-253.241608</td>\n",
       "      <td>-196.550934</td>\n",
       "      <td>first</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-166.223312</td>\n",
       "      <td>-377.834686</td>\n",
       "      <td>sentence</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   TSNE component 1  TSNE component 2      word\n",
       "0        266.298431         23.642027      this\n",
       "1        132.896149       -161.362015        is\n",
       "2       -329.322601        -13.235461       the\n",
       "3       -253.241608       -196.550934     first\n",
       "4       -166.223312       -377.834686  sentence"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(tsne,columns = ['TSNE component 1', 'TSNE component 2'])\n",
    "df['word'] = model.wv.vocab.keys()\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "def tsne_plot(model):\n",
    "    \"Creates and TSNE model and plots it\"\n",
    "    labels = []\n",
    "    tokens = []\n",
    "\n",
    "    for word in model.wv.vocab:\n",
    "        tokens.append(model[word])\n",
    "        labels.append(word)\n",
    "    \n",
    "    tsne_model = TSNE(perplexity=40, n_components=2, init='pca', n_iter=2500, random_state=23)\n",
    "    new_values = tsne_model.fit_transform(tokens)\n",
    "\n",
    "    x = []\n",
    "    y = []\n",
    "    for value in new_values:\n",
    "        x.append(value[0])\n",
    "        y.append(value[1])\n",
    "        \n",
    "    plt.figure(figsize=(10, 5)) \n",
    "    for i in range(len(x)):\n",
    "        plt.scatter(x[i],y[i])\n",
    "        plt.annotate(labels[i],\n",
    "                     xy=(x[i], y[i]),\n",
    "                     xytext=(5, 2),\n",
    "                     textcoords='offset points',\n",
    "                     ha='right',\n",
    "                     va='bottom')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "tsne_plot(model)"
   ]
  }
 ],
 "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}