{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll start by using the [markovify](https://github.com/jsvine/markovify/) library to make some individual sentences in the style of Jane Austen.  These will be the basis for generating a stream of synthetic documents."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import markovify\n",
    "import codecs\n",
    "import random\n",
    "\n",
    "# Markovify uses a single random generator -- notebooks using it will thus \n",
    "# only be reproducible if you set a random seed before each cell using markovify\n",
    "random.seed(0xbaff1ed)\n",
    "\n",
    "with codecs.open(\"data/austen.txt\", \"r\", \"cp1252\") as f:\n",
    "    text = f.read()\n",
    "\n",
    "austen_model = markovify.Text(text, retain_original=False, state_size=3)\n",
    "\n",
    "for i in range(10):\n",
    "    print(austen_model.make_short_sentence(200))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Constructing single sentences is interesting, but we'd really rather construct larger documents. Here we'll construct a series of documents that have, on average, five sentences."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "from scipy.stats import poisson\n",
    "import numpy as np\n",
    "\n",
    "def make_basic_documents(sentence_count=5, document_count=1, model=austen_model, seed=None):\n",
    "    def shortsentence(ct):\n",
    "        return \" \".join([model.make_short_sentence(200) for _ in range(ct + 1)])\n",
    "\n",
    "    if seed is not None:\n",
    "        # seed both the Python generator and the NumPy one used by SciPy\n",
    "        random.seed(seed)\n",
    "        np.random.seed(seed)\n",
    "    \n",
    "    return [shortsentence(ct) for ct in poisson.rvs(sentence_count, size=document_count)]\n",
    "\n",
    "for doc in make_basic_documents(5, 10, seed=0xdecaf):\n",
    "    print(doc)\n",
    "    print(\"\\n###\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We're going to use the Austen model as the main basis for _legitimate messages_ in our sample data set.  For the _spam messages_, we'll train two Markov models on positive and negative product reviews (taken from the [public-domain Amazon fine foods reviews dataset on Kaggle](https://www.kaggle.com/snap/amazon-fine-food-reviews/)).  We'll combine the models from these sources in different proportions so that all words are _possible_ in certain kinds of messages but some words are _more likely_ in legitimate messages or in spam."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gzip\n",
    "\n",
    "def train_markov_gz(fn):\n",
    "    \"\"\" trains a Markov model on gzipped text data \"\"\"\n",
    "    with gzip.open(fn, \"rt\", encoding=\"utf-8\") as f:\n",
    "        text = f.read()\n",
    "    return markovify.Text(text, retain_original=False, state_size=3)\n",
    "\n",
    "negative_model = train_markov_gz(\"data/reviews-1.txt.gz\")\n",
    "positive_model = train_markov_gz(\"data/reviews-5-100k.txt.gz\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can combine these models with relative weights, but this yields somewhat unusual results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "legitimate_model = markovify.combine([austen_model, negative_model, positive_model], [196, 2, 2])\n",
    "spam_model = markovify.combine([austen_model, negative_model, positive_model], [3, 30, 40])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# seed both the Python generator and the NumPy one used by SciPy\n",
    "random.seed(0xc0ffee)\n",
    "np.random.seed(0xc0ffee)\n",
    "\n",
    "for s in make_basic_documents(5, 20, legitimate_model):\n",
    "    print(s)\n",
    "    print(\"\\n###\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "random.seed(0xf00)\n",
    "np.random.seed(0xf00)\n",
    "\n",
    "for s in make_basic_documents(5, 20, spam_model):\n",
    "    print(s)\n",
    "    print(\"\\n###\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can then generate some example documents and save them to a file for use in the next notebook.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "pd.set_option(\"io.parquet.engine\", \"pyarrow\")\n",
    "\n",
    "random.seed(0xda7aba5e)\n",
    "np.random.seed(0xda7aba5e)\n",
    "\n",
    "df = pd.DataFrame(columns=[\"label\", \"text\"], dtype=np.dtype(str))\n",
    "\n",
    "mean_sentences_per_example = 5\n",
    "examples_per_class = 20000\n",
    "\n",
    "for (label, model) in [(\"legitimate\", legitimate_model), (\"spam\", spam_model)]:\n",
    "    docs = [{\"label\" : label, \"text\" : txt} for txt in make_basic_documents(mean_sentences_per_example, examples_per_class, model)]\n",
    "    df = pd.concat([df, pd.DataFrame(docs)])\n",
    "\n",
    "df[\"text\"] = df[\"text\"].astype(\"str\")\n",
    "df[\"label\"] = df[\"label\"].astype(\"category\")\n",
    "df.reset_index().to_parquet(\"data/training.parquet\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's go to [the next notebook](01-vectors-and-visualization.ipynb) now!"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.6",
   "language": "python",
   "name": "jupyter"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}