Explore different methods for tokenizing

explore_tokenizers.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Explore different tokenizers: both quality and performance\n",
    "====\n",
    "\n",
    "HT:\n",
    "- https://github.com/norvig/pytudes/blob/master/How%20to%20Do%20Things%20with%20Words.ipynb\n",
    "- https://www.oreilly.com/learning/how-can-i-tokenize-a-sentence-with-python"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "reset -fs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from urllib import request"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get a reasonable chuck of text data to test speed\n",
    "url = \"https://www.wolframcloud.com/objects/fd2da57e-2af0-4114-9d08-f45c062389d4\" # Leaves of Grass by Whitman\n",
    "text = request.urlopen(url).read().decode(\"utf8\") # NOTE: Convert from bytes to str"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'LEAVES OF GRASS\\n\\nBy Walt Whitman\\n\\n\\n\\n     Come, said my soul,\\n     Such verses for my Body let us write, (for we are one,)\\n     That should I after return,\\n     Or, long, long hence, in other spheres,\\n     '"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text[:205]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Tests suite\n",
    "def tokenize_example_text(tokenizer_function_under_test):\n",
    "    text = \"This is: A test, 1, 2, 3, this-is.\"\n",
    "    print(f\"Orginal:\\t{text}\")\n",
    "    print(\" \")\n",
    "    print(\"Tokenized: \", end=\"\\t\")\n",
    "    print(*tokenizer_function_under_test(text), sep=\"\\n\\t\\t\")\n",
    "    \n",
    "def evaluate_tokenizer(tokenizer_function_under_test, text=text):\n",
    "    tokenizer_function_under_test(text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Built-in string method\n",
    "-----"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define tokenizer function\n",
    "tokens_str_split = str.split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Orginal:\tThis is: A test, 1, 2, 3, this-is.\n",
      " \n",
      "Tokenized: \tThis\n",
      "\t\tis:\n",
      "\t\tA\n",
      "\t\ttest,\n",
      "\t\t1,\n",
      "\t\t2,\n",
      "\t\t3,\n",
      "\t\tthis-is.\n"
     ]
    }
   ],
   "source": [
    "# Visually inspect results for quality\n",
    "tokenize_example_text(tokens_str_split)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate performance speed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12.7 ms ± 2.18 ms per loop (mean ± std. dev. of 7 runs, 3 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit -n3\n",
    "\n",
    "evaluate_tokenizer(tokens_str_split)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Regex methods\n",
    "----"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "import re"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define tokenizer function\n",
    "def tokens_regex_restricted(text):\n",
    "    \"List all the word tokens in a text.\"\n",
    "    return re.findall('[\\w]+', text.lower())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Orginal:\tThis is: A test, 1, 2, 3, this-is.\n",
      " \n",
      "Tokenized: \tthis\n",
      "\t\tis\n",
      "\t\ta\n",
      "\t\ttest\n",
      "\t\t1\n",
      "\t\t2\n",
      "\t\t3\n",
      "\t\tthis\n",
      "\t\tis\n"
     ]
    }
   ],
   "source": [
    "# Visually inspect results for quality\n",
    "tokenize_example_text(tokens_regex_restricted)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate performance speed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "55.4 ms ± 7.81 ms per loop (mean ± std. dev. of 7 runs, 3 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit -n3\n",
    "\n",
    "evaluate_tokenizer(tokens_regex_restricted)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define tokenizer function\n",
    "def tokens_regex_improved(text):\n",
    "    \"List all the word tokens (consecutive letters) in a text. Normalize to lowercase.\"\n",
    "    return re.findall('[a-z]+', text.lower())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Orginal:\tThis is: A test, 1, 2, 3, this-is.\n",
      " \n",
      "Tokenized: \tthis\n",
      "\t\tis\n",
      "\t\ta\n",
      "\t\ttest\n",
      "\t\tthis\n",
      "\t\tis\n"
     ]
    }
   ],
   "source": [
    "# Visually inspect results for quality\n",
    "tokenize_example_text(tokens_regex_improved)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate performance speed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "46.9 ms ± 2.2 ms per loop (mean ± std. dev. of 7 runs, 3 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit -n3\n",
    "\n",
    "evaluate_tokenizer(tokens_regex_improved)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Python packages\n",
    "----\n",
    "\n",
    "### nltk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# NOTE: nltk does not support 3.7 ☹\n",
    "\n",
    "# # Import tokenizer function\n",
    "# from nltk.tokenize import word_tokenize as tokens_nltk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# # Visually inspect results for quality\n",
    "# tokenize_example_text(tokens_nltk)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate performance speed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %%timeit -n3\n",
    "\n",
    "# evaluate_tokenizer(tokens_nltk) # Takes about 1 second in wall time"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### spacCy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "import spacy\n",
    "\n",
    "nlp = spacy.load('en_core_web_sm') # Import language model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define helper function\n",
    "def tokens_spacy(text):\n",
    "    doc = nlp.tokenizer(text)\n",
    "    return [token for token in doc]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Orginal:\tThis is: A test, 1, 2, 3, this-is.\n",
      " \n",
      "Tokenized: \tThis\n",
      "\t\tis\n",
      "\t\t:\n",
      "\t\tA\n",
      "\t\ttest\n",
      "\t\t,\n",
      "\t\t1\n",
      "\t\t,\n",
      "\t\t2\n",
      "\t\t,\n",
      "\t\t3\n",
      "\t\t,\n",
      "\t\tthis\n",
      "\t\t-\n",
      "\t\tis\n",
      "\t\t.\n"
     ]
    }
   ],
   "source": [
    "# Visually inspect results for quality\n",
    "tokenize_example_text(tokens_spacy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate performance speed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4.94 s ± 135 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit -n1\n",
    "\n",
    "evaluate_tokenizer(tokens_spacy) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br>\n",
    "<br> \n",
    "<br>\n",
    "\n",
    "----"
   ]
  }
 ],
 "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}