1. Experiment.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "Bad key \"text.kerning_factor\" on line 4 in\n",
      "/home/gagan/miniconda3/envs/nlp/lib/python3.7/site-packages/matplotlib/mpl-data/stylelib/_classic_test_patch.mplstyle.\n",
      "You probably need to get an updated matplotlibrc file from\n",
      "https://github.com/matplotlib/matplotlib/blob/v3.1.3/matplotlibrc.template\n",
      "or from the matplotlib source distribution\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_file_path = 'data/big.txt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "accepted_chars = 'abcdefghijklmnopqrstuvwxyz '\n",
    "len_accepted_chars = len(accepted_chars)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'a': 0,\n",
       " 'b': 1,\n",
       " 'c': 2,\n",
       " 'd': 3,\n",
       " 'e': 4,\n",
       " 'f': 5,\n",
       " 'g': 6,\n",
       " 'h': 7,\n",
       " 'i': 8,\n",
       " 'j': 9,\n",
       " 'k': 10,\n",
       " 'l': 11,\n",
       " 'm': 12,\n",
       " 'n': 13,\n",
       " 'o': 14,\n",
       " 'p': 15,\n",
       " 'q': 16,\n",
       " 'r': 17,\n",
       " 's': 18,\n",
       " 't': 19,\n",
       " 'u': 20,\n",
       " 'v': 21,\n",
       " 'w': 22,\n",
       " 'x': 23,\n",
       " 'y': 24,\n",
       " 'z': 25,\n",
       " ' ': 26}"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "char_to_idx = {c: idx for idx, c in enumerate(accepted_chars)}; char_to_idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def clean_and_tokenize(text):\n",
    "    return [char.lower() for char in text if char.lower() in accepted_chars]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['t', 'h', 'i', 's', ' ', 's', 'h', 'o', 'u', 'l', 'd', ' ', 'b', 'e']"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clean_and_tokenize(\"this should be?\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def ngram(n, txt):\n",
    "    tokenized_txt = clean_and_tokenize(txt)\n",
    "    for idx in range(len(tokenized_txt)-n+1):\n",
    "        yield tokenized_txt[idx:idx+n]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<generator object ngram at 0x7fa9bf0f9250>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ngram(2, \"hello\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.],\n",
       "       [10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10., 10.,\n",
       "        10.]])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probabilites = np.ones((27,27))*10; probabilites"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fa9bf10eb10>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(10,8))\n",
    "sns.heatmap(probabilites, xticklabels=char_to_idx.keys(), yticklabels=char_to_idx.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(probabilites) == len(char_to_idx)\n",
    "\n",
    "# if they don't match then something is wrong"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "45542it [00:03, 12031.93it/s]\n"
     ]
    }
   ],
   "source": [
    "for line in tqdm(open(train_file_path, 'r')):\n",
    "    for fc, sc in ngram(2, line):\n",
    "        probabilites[char_to_idx[fc]][char_to_idx[sc]] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Probability of ab = 3602.0\n",
      "Probability of aa = 44.0\n"
     ]
    }
   ],
   "source": [
    "# let's look at some probabilies\n",
    "\n",
    "print(f\"Probability of ab = {probabilites[char_to_idx['a']][char_to_idx['b']]}\")\n",
    "print(f\"Probability of aa = {probabilites[char_to_idx['a']][char_to_idx['a']]}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "for idx, row in enumerate(probabilites):\n",
    "    s = np.sum(row)\n",
    "    pb = np.log(row/s)\n",
    "    probabilites[idx] = pb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fa9bc401490>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(10,8))\n",
    "sns.heatmap(probabilites, xticklabels=char_to_idx.keys(), yticklabels=char_to_idx.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(text, probabilities, char_to_idx):\n",
    "    pb = 0\n",
    "    for fc, sc in ngram(2, text):\n",
    "        pb += probabilities[char_to_idx[fc]][char_to_idx[sc]]\n",
    "    print(pb)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-5.421759346590587\n"
     ]
    }
   ],
   "source": [
    "predict(\"yes\", probabilites, char_to_idx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-10.888981769066623\n"
     ]
    }
   ],
   "source": [
    "predict(\"gagan\", probabilites, char_to_idx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.7.7 64-bit ('nlp': conda)",
   "language": "python",
   "name": "python37764bitnlpconda023c561587e543c88e55ee37daf8a2ed"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}