notebooks/nlp with tensorflow 2 - tokenizer and sequencer.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentences = [\n",
    "    'I love my dog',\n",
    "    'I love my cat',\n",
    "    'You love dog!',\n",
    "    'Do you think my dog is amazing?']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'<OOV>': 1, 'love': 2, 'my': 3, 'dog': 4, 'i': 5, 'you': 6, 'cat': 7, 'do': 8, 'think': 9, 'is': 10, 'amazing': 11}\n"
     ]
    }
   ],
   "source": [
    "# num_words keeps only 100 words\n",
    "tokenizer = Tokenizer(num_words = 100, oov_token=\"<OOV>\")\n",
    "\n",
    "# here tokenize the sentences using tokenizer\n",
    "tokenizer.fit_on_texts(sentences)\n",
    "\n",
    "#word index has the word and its index\n",
    "# notice how tokenizer ignored symbols\n",
    "word_index = tokenizer.word_index\n",
    "\n",
    "print(word_index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "sequence of numbers in correct order makes up a sentence, this sequence will be used to train neural networks<br>\n",
    "tokenizer.text_to_sequences(sentence) does the job us and turns sentences in to sequence of tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "sequences = tokenizer.texts_to_sequences(sentences)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'<OOV>': 1, 'love': 2, 'my': 3, 'dog': 4, 'i': 5, 'you': 6, 'cat': 7, 'do': 8, 'think': 9, 'is': 10, 'amazing': 11}\n"
     ]
    }
   ],
   "source": [
    "print(word_index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[5, 2, 3, 4], [5, 2, 3, 7], [6, 2, 4], [8, 6, 9, 3, 4, 10, 11]]\n"
     ]
    }
   ],
   "source": [
    "print(sequences)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### pad_sequences turns sequences to padded sequences "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "padded = pad_sequences(sequences, padding='post', truncating='post',\n",
    "                      maxlen=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[5 2 3 4 0]\n",
      " [5 2 3 7 0]\n",
      " [6 2 4 0 0]\n",
      " [8 6 9 3 4]]\n"
     ]
    }
   ],
   "source": [
    "print(padded)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### what happens when we need to classify words which the neural network hasn't learn?\n",
    "lets try below"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_data = [\n",
    "    'i really love my dog',\n",
    "    'my dog loves my mantee'\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_seq = tokenizer.texts_to_sequences(test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[5, 1, 2, 3, 4], [3, 4, 1, 3, 1]]\n"
     ]
    }
   ],
   "source": [
    "print(test_seq)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note: the tokenizer used the word < OOV > = 1 for out of vocabulary words which weren't in the traiing of the tokenizer<br>\n",
    "Otherwise if the oov wasn't specified in the tokenizer then the tokenizer would simply have ignoed the word"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### allow out of index < OOV > for words that aren't used to train the tokenizer   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**pad_sequeces** can be used to pad the sequence and make the sentences of equal length, putting zeros where the word isn't present"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**maxlen**, **truncating** in pad_sequences maxlen and truncating can be used to keep the sentence of some particular length, it will truncate the tail if truncating is set to 'post'"
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}