slides_preprocessing.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Playground for Preprocessing Slides\n",
    "\n",
    "- Stephen W. Thomas\n",
    "- Used for MMAI 891 and MMA 865"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2018-12-12 14:32:53.428780\n"
     ]
    }
   ],
   "source": [
    "import datetime\n",
    "print(datetime.datetime.now())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns \n",
    "\n",
    "import re\n",
    "import string\n",
    "import unicodedata\n",
    "\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "from nltk import word_tokenize\n",
    "from nltk import word_tokenize, sent_tokenize\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.stem import LancasterStemmer, WordNetLemmatizer\n",
    "\n",
    "from bs4 import BeautifulSoup\n",
    "\n",
    "import os\n",
    "\n",
    "from IPython.core.interactiveshell import InteractiveShell\n",
    "InteractiveShell.ast_node_interactivity = \"all\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Case Normalization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'hi, my name is steve.'"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'i work at the u.n. in nyc in the usa.'"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\"Hi, my name is Steve.\".lower()\n",
    "'I work at the U.N. in NYC in the USA.'.lower()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tokenization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For simple tokenzation, the NLTK docs recommend to just use `.split()`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Hi,', 'my', 'name', 'is', 'Steve.']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[\"Let's\", 'go', 'to', 'the', 'U.N.', 'in', 'NYC!!']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "['The',\n",
       " 'ever-popular',\n",
       " 'San',\n",
       " 'Fransisco',\n",
       " 'love',\n",
       " 'letter',\n",
       " 'arrived',\n",
       " 'on',\n",
       " 'March',\n",
       " '11,',\n",
       " '2019.']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "['stephen.thomas@queensu.ca']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "['(613)', '453-6162']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\"Hi, my name is Steve.\".split()\n",
    "\"Let's go to the U.N. in NYC!!\".split()\n",
    "\"The ever-popular San Fransisco love letter arrived on March 11, 2019.\".split()\n",
    "\"stephen.thomas@queensu.ca\".split()\n",
    "\"(613) 453-6162\".split()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Can also use `nltk`'s `word_tokenize()` function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Hi', ',', 'my', 'name', 'is', 'Steve', '.']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "['Let', \"'s\", 'go', 'to', 'the', 'U.N.', 'in', 'NYC', '!', '!']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "['The',\n",
       " 'ever-popular',\n",
       " 'San',\n",
       " 'Fransisco',\n",
       " 'love',\n",
       " 'letter',\n",
       " 'arrived',\n",
       " 'on',\n",
       " 'March',\n",
       " '11',\n",
       " ',',\n",
       " '2019',\n",
       " '.']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "['stephen.thomas', '@', 'queensu.ca']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "['(', '613', ')', '453-6162']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from nltk import word_tokenize\n",
    "word_tokenize(\"Hi, my name is Steve.\")\n",
    "word_tokenize(\"Let's go to the U.N. in NYC!!\")\n",
    "word_tokenize(\"The ever-popular San Fransisco love letter arrived on March 11, 2019.\")\n",
    "word_tokenize(\"stephen.thomas@queensu.ca\")\n",
    "word_tokenize(\"(613) 453-6162\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`nltk` even has a tokenizer for tweets!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['This',\n",
       " 'is',\n",
       " 'a',\n",
       " 'cooool',\n",
       " '#dummysmiley',\n",
       " ':',\n",
       " ':-)',\n",
       " ':-P',\n",
       " '<3',\n",
       " 'and',\n",
       " 'some',\n",
       " 'arrows',\n",
       " '<',\n",
       " '>',\n",
       " '->',\n",
       " '<--']"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from nltk.tokenize import TweetTokenizer\n",
    "tknzr = TweetTokenizer()\n",
    "s0 = \"This is a cooool #dummysmiley: :-) :-P <3 and some arrows < > -> <--\"\n",
    "tknzr.tokenize(s0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# N-Grams"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Hey', 'there')\n",
      "('there', ',')\n",
      "(',', 'I')\n",
      "('I', 'am')\n",
      "('am', 'awesome')\n",
      "('awesome', '.')\n",
      "('Hey', 'there', ',')\n",
      "('there', ',', 'I')\n",
      "(',', 'I', 'am')\n",
      "('I', 'am', 'awesome')\n",
      "('am', 'awesome', '.')\n"
     ]
    }
   ],
   "source": [
    "from nltk.util import ngrams\n",
    "tokens = nltk.word_tokenize(\"Hey there, I am awesome.\")\n",
    "\n",
    "grams = ngrams(tokens,2)\n",
    "for grams in grams:\n",
    "  print(grams)\n",
    "\n",
    "grams = ngrams(tokens,3)\n",
    "for grams in grams:\n",
    "  print(grams)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Removing Characters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Hey Lets go to the bar'"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import re\n",
    "s = \"Hey! Let's go to the bar...\"\n",
    "re.sub(r'[^\\w\\s]', '', s)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Abcdeacoauu'"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import unidecode\n",
    "s = u'Ábcdêãçoàúü'\n",
    "unidecode.unidecode(s)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Removing Numbers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'There are only  classes left.'"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import re\n",
    "s = \"There are only 4 classes left.\"\n",
    "re.sub(r'\\d+', '', s)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Regular Expressions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Alejandrina has a small store in her house where she sells basic products. ',\n",
       " 'Nancy works as a saleswoman, her main product being used clothing.',\n",
       " 'Marjorie sells lunch to companies and construction workers.  Kolb.']"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sentences = [\n",
    "  \"Alejandrina has a small store in her house where she sells basic products. Translated from Spanish by Jennifer Day, Kiva Volunteer.\",\n",
    "  \"Nancy works as a saleswoman, her main product being used clothing.Translated from Spanish by Kiva Volunteer, Kristin Connor.\",\n",
    "  \"Marjorie sells lunch to companies and construction workers. Translated by Ramn F. Kolb.\"]\n",
    "\n",
    "[re.sub(r'Translated[^\\.]+\\.', '', s) for s in sentences]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Stemming and Lemmatization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "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>Orig</th>\n",
       "      <th>Porter</th>\n",
       "      <th>Snowball</th>\n",
       "      <th>Lancaster</th>\n",
       "      <th>WordNetLemmatizer</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>caresses</td>\n",
       "      <td>caress</td>\n",
       "      <td>caress</td>\n",
       "      <td>caress</td>\n",
       "      <td>caress</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>flies</td>\n",
       "      <td>fli</td>\n",
       "      <td>fli</td>\n",
       "      <td>fli</td>\n",
       "      <td>fly</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>dies</td>\n",
       "      <td>die</td>\n",
       "      <td>die</td>\n",
       "      <td>die</td>\n",
       "      <td>dy</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>mules</td>\n",
       "      <td>mule</td>\n",
       "      <td>mule</td>\n",
       "      <td>mul</td>\n",
       "      <td>mule</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>denied</td>\n",
       "      <td>deni</td>\n",
       "      <td>deni</td>\n",
       "      <td>deny</td>\n",
       "      <td>denied</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>died</td>\n",
       "      <td>die</td>\n",
       "      <td>die</td>\n",
       "      <td>died</td>\n",
       "      <td>died</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>agreed</td>\n",
       "      <td>agre</td>\n",
       "      <td>agre</td>\n",
       "      <td>agree</td>\n",
       "      <td>agreed</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>owned</td>\n",
       "      <td>own</td>\n",
       "      <td>own</td>\n",
       "      <td>own</td>\n",
       "      <td>owned</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>humbled</td>\n",
       "      <td>humbl</td>\n",
       "      <td>humbl</td>\n",
       "      <td>humbl</td>\n",
       "      <td>humbled</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>sized</td>\n",
       "      <td>size</td>\n",
       "      <td>size</td>\n",
       "      <td>siz</td>\n",
       "      <td>sized</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>meeting</td>\n",
       "      <td>meet</td>\n",
       "      <td>meet</td>\n",
       "      <td>meet</td>\n",
       "      <td>meeting</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>stating</td>\n",
       "      <td>state</td>\n",
       "      <td>state</td>\n",
       "      <td>stat</td>\n",
       "      <td>stating</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>itemization</td>\n",
       "      <td>item</td>\n",
       "      <td>item</td>\n",
       "      <td>item</td>\n",
       "      <td>itemization</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>sensational</td>\n",
       "      <td>sensat</td>\n",
       "      <td>sensat</td>\n",
       "      <td>sens</td>\n",
       "      <td>sensational</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>traditional</td>\n",
       "      <td>tradit</td>\n",
       "      <td>tradit</td>\n",
       "      <td>tradit</td>\n",
       "      <td>traditional</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>reference</td>\n",
       "      <td>refer</td>\n",
       "      <td>refer</td>\n",
       "      <td>ref</td>\n",
       "      <td>reference</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>colonizer</td>\n",
       "      <td>colon</td>\n",
       "      <td>colon</td>\n",
       "      <td>colon</td>\n",
       "      <td>colonizer</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>plotted</td>\n",
       "      <td>plot</td>\n",
       "      <td>plot</td>\n",
       "      <td>plot</td>\n",
       "      <td>plotted</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>generously</td>\n",
       "      <td>gener</td>\n",
       "      <td>generous</td>\n",
       "      <td>gen</td>\n",
       "      <td>generously</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>crying</td>\n",
       "      <td>cri</td>\n",
       "      <td>cri</td>\n",
       "      <td>cry</td>\n",
       "      <td>cry</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           Orig  Porter  Snowball Lancaster WordNetLemmatizer\n",
       "0      caresses  caress    caress    caress            caress\n",
       "1         flies     fli       fli       fli               fly\n",
       "2          dies     die       die       die                dy\n",
       "3         mules    mule      mule       mul              mule\n",
       "4        denied    deni      deni      deny            denied\n",
       "5          died     die       die      died              died\n",
       "6        agreed    agre      agre     agree            agreed\n",
       "7         owned     own       own       own             owned\n",
       "8       humbled   humbl     humbl     humbl           humbled\n",
       "9         sized    size      size       siz             sized\n",
       "10      meeting    meet      meet      meet           meeting\n",
       "11      stating   state     state      stat           stating\n",
       "12  itemization    item      item      item       itemization\n",
       "13  sensational  sensat    sensat      sens       sensational\n",
       "14  traditional  tradit    tradit    tradit       traditional\n",
       "15    reference   refer     refer       ref         reference\n",
       "16    colonizer   colon     colon     colon         colonizer\n",
       "17      plotted    plot      plot      plot           plotted\n",
       "18   generously   gener  generous       gen        generously\n",
       "19       crying     cri       cri       cry               cry"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from nltk.stem.porter import PorterStemmer\n",
    "from nltk.stem.snowball import SnowballStemmer\n",
    "from nltk.stem.lancaster import LancasterStemmer\n",
    "from nltk.stem import WordNetLemmatizer\n",
    "\n",
    "\n",
    "words = ['caresses', 'flies', 'dies', 'mules', 'denied',\n",
    "            'died', 'agreed', 'owned', 'humbled', 'sized',\n",
    "            'meeting', 'stating', 'itemization',\n",
    "            'sensational', 'traditional', 'reference', 'colonizer',\n",
    "            'plotted', 'generously', 'crying']\n",
    "\n",
    "res = pd.DataFrame()\n",
    "res = res.assign(Orig=words,\n",
    "                 Porter=[PorterStemmer().stem(w) for w in words],\n",
    "                 Snowball=[SnowballStemmer('english').stem(w) for w in words],\n",
    "                 Lancaster=[LancasterStemmer().stem(w) for w in words],\n",
    "                 WordNetLemmatizer=[WordNetLemmatizer().lemmatize(w) for w in words])\n",
    "res"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Spell Checking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 47 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('car', 0.5), ('czar', 0.42857142857142855), ('saar', 0.07142857142857142)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('message', 0.6216216216216216), ('massage', 0.3783783783783784)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('service', 0.9253112033195021), ('survive', 0.07468879668049792)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('the', 0.8653201565642368),\n",
       " ('he', 0.13408515883485067),\n",
       " ('ate', 0.00022706139307570876),\n",
       " ('hate', 0.0002162489457863893),\n",
       " ('hue', 0.00012974936747183358),\n",
       " ('te', 1.0812447289319465e-05),\n",
       " ('htm', 1.0812447289319465e-05)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('whisky', 0.9333333333333333), ('risky', 0.06666666666666667)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from pattern.en import suggest\n",
    "\n",
    "%time suggest(u'caar')\n",
    "suggest(u'mussage')\n",
    "suggest(u'survice')\n",
    "suggest(u'hte')\n",
    "suggest(u\"wisky\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Stopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['name', 'steve', ',', 'good', 'chef', '.']"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from nltk.corpus import stopwords\n",
    "tokens = nltk.word_tokenize(\"My name is Steve, and I am a good chef.\".lower())\n",
    "[t for t in tokens if t not in stopwords.words('english')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['a', 'about', 'above', 'after', 'again', 'against', 'ain', 'all', 'am', 'an',\n",
      " 'and', 'any', 'are', 'aren', \"aren't\", 'as', 'at', 'be', 'because', 'been',\n",
      " 'before', 'being', 'below', 'between', 'both', 'but', 'by', 'can', 'couldn',\n",
      " \"couldn't\", 'd', 'did', 'didn', \"didn't\", 'do', 'does', 'doesn', \"doesn't\",\n",
      " 'doing', 'don', \"don't\", 'down', 'during', 'each', 'few', 'for', 'from',\n",
      " 'further', 'had', 'hadn', \"hadn't\", 'has', 'hasn', \"hasn't\", 'have', 'haven',\n",
      " \"haven't\", 'having', 'he', 'her', 'here', 'hers', 'herself', 'him', 'himself',\n",
      " 'his', 'how', 'i', 'if', 'in', 'into', 'is', 'isn', \"isn't\", 'it', \"it's\",\n",
      " 'its', 'itself', 'just', 'll', 'm', 'ma', 'me', 'mightn', \"mightn't\", 'more',\n",
      " 'most', 'mustn', \"mustn't\", 'my', 'myself', 'needn', \"needn't\", 'no', 'nor',\n",
      " 'not', 'now', 'o', 'of', 'off', 'on', 'once', 'only', 'or', 'other', 'our',\n",
      " 'ours', 'ourselves', 'out', 'over', 'own', 're', 's', 'same', 'shan', \"shan't\",\n",
      " 'she', \"she's\", 'should', \"should've\", 'shouldn', \"shouldn't\", 'so', 'some',\n",
      " 'such', 't', 'than', 'that', \"that'll\", 'the', 'their', 'theirs', 'them',\n",
      " 'themselves', 'then', 'there', 'these', 'they', 'this', 'those', 'through',\n",
      " 'to', 'too', 'under', 'until', 'up', 've', 'very', 'was', 'wasn', \"wasn't\",\n",
      " 'we', 'were', 'weren', \"weren't\", 'what', 'when', 'where', 'which', 'while',\n",
      " 'who', 'whom', 'why', 'will', 'with', 'won', \"won't\", 'wouldn', \"wouldn't\",\n",
      " 'y', 'you', \"you'd\", \"you'll\", \"you're\", \"you've\", 'your', 'yours', 'yourself',\n",
      " 'yourselves']\n"
     ]
    }
   ],
   "source": [
    "import pprint\n",
    "pp = pprint.PrettyPrinter(width=80, compact=True)\n",
    "pp.pprint(sorted(stopwords.words('english')))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fuzzy Matching"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Queens University'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[(\"Queen's University\", 97),\n",
       " ('University of Toronto', 70),\n",
       " ('McGill University', 70)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "\"Queen's\""
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[(\"Queen's University\", 90),\n",
       " ('University of Toronto', 37),\n",
       " ('McGill University', 26)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'Queens Univrsity'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[(\"Queen's University\", 94),\n",
       " ('McGill University', 61),\n",
       " ('University of Toronto', 56)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'U of T'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('University of Toronto', 86),\n",
       " ('University Of Waterloo', 86),\n",
       " (\"Queen's University\", 43)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'Toronto'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('University of Toronto', 90),\n",
       " ('University Of Waterloo', 51),\n",
       " (\"Queen's University\", 26)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'Rotman'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('University of Toronto', 45),\n",
       " (\"Queen's University\", 30),\n",
       " ('McGill University', 30)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'Mcgill'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('McGill University', 90),\n",
       " (\"Queen's University\", 15),\n",
       " ('University of Toronto', 15)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'McGill U.'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('McGill University', 90),\n",
       " (\"Queen's University\", 22),\n",
       " ('University of Toronto', 22)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'Waterloo'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('University Of Waterloo', 90),\n",
       " ('University of Toronto', 45),\n",
       " (\"Queen's University\", 22)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'Waterloop University'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "[('University Of Waterloo', 86),\n",
       " ('McGill University', 70),\n",
       " ('University of Toronto', 69)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from fuzzywuzzy import fuzz\n",
    "from fuzzywuzzy import process\n",
    "\n",
    "choices = ['Queen\\'s University', 'University of Toronto', 'McGill University', 'University Of Waterloo']\n",
    "\n",
    "typos = ['Queens University', 'Queen\\'s', 'Queens Univrsity',\n",
    "                         'U of T', 'Toronto', 'Rotman',\n",
    "                         'Mcgill', 'McGill U.',\n",
    "                         'Waterloo', 'Waterloop University']\n",
    "\n",
    "for typo in typos:\n",
    "    typo\n",
    "    process.extract(typo, choices, limit=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Putting it all together"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Simple Example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3, 6)\n"
     ]
    },
    {
     "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>brown</th>\n",
       "      <th>brown furry</th>\n",
       "      <th>furry</th>\n",
       "      <th>matrix</th>\n",
       "      <th>quick</th>\n",
       "      <th>the quick</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   brown  brown furry  furry  matrix  quick  the quick\n",
       "0      2            1      2       0      1          1\n",
       "1      1            1      1       1      0          0\n",
       "2      0            0      0       1      1          1"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "\n",
    "myText = [\"the quick brown furry fox jumped over a second furry brown fox\",\n",
    "              \"the sparse brown furry matrix\",\n",
    "              \"the quick matrix\"]\n",
    "\n",
    "\n",
    "tf_vectorizer = CountVectorizer(min_df=.5, max_df=.75, ngram_range=[1,3])\n",
    "dtm_tf = tf_vectorizer.fit_transform(myText)\n",
    "print(dtm_tf.shape)\n",
    "\n",
    "df = pd.DataFrame(dtm_tf.toarray(), columns=tf_vectorizer.get_feature_names())\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Bigger Example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 206 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['status', 'sector', 'en', 'country', 'gender', 'loan_amount', 'nonpayment']"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(7882, 7)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>status</th>\n",
       "      <th>sector</th>\n",
       "      <th>en</th>\n",
       "      <th>country</th>\n",
       "      <th>gender</th>\n",
       "      <th>loan_amount</th>\n",
       "      <th>nonpayment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Agriculture</td>\n",
       "      <td>&lt;i&gt;This description is written by Rowland Amul...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>&lt;i&gt;This description is written by Rowland Amul...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>&lt;i&gt;This description was written by Richard Maz...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Services</td>\n",
       "      <td>&lt;i&gt;This description was written by Rowland Amu...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Construction</td>\n",
       "      <td>&lt;i&gt;This description was written by Rowland Amu...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      status        sector                                                 en  \\\n",
       "0  defaulted   Agriculture  <i>This description is written by Rowland Amul...   \n",
       "1  defaulted          Food  <i>This description is written by Rowland Amul...   \n",
       "2  defaulted          Food  <i>This description was written by Richard Maz...   \n",
       "3  defaulted      Services  <i>This description was written by Rowland Amu...   \n",
       "4  defaulted  Construction  <i>This description was written by Rowland Amu...   \n",
       "\n",
       "  country gender  loan_amount nonpayment  \n",
       "0   Kenya      M        500.0     lender  \n",
       "1   Kenya      F        500.0     lender  \n",
       "2   Kenya      M        500.0     lender  \n",
       "3   Kenya      F        500.0     lender  \n",
       "4   Kenya      M        500.0     lender  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>status</th>\n",
       "      <th>sector</th>\n",
       "      <th>en</th>\n",
       "      <th>country</th>\n",
       "      <th>gender</th>\n",
       "      <th>loan_amount</th>\n",
       "      <th>nonpayment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>7983</th>\n",
       "      <td>paid</td>\n",
       "      <td>Clothing</td>\n",
       "      <td>Rosa is a member of the 3 de Mayo Communal Ban...</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>1050.0</td>\n",
       "      <td>lender</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7984</th>\n",
       "      <td>paid</td>\n",
       "      <td>Agriculture</td>\n",
       "      <td>The communal bank, \"Cumbe,\" is located on the ...</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>2650.0</td>\n",
       "      <td>partner</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7985</th>\n",
       "      <td>paid</td>\n",
       "      <td>Food</td>\n",
       "      <td>Gladys Burgos Macas is a member of the 3 de Ma...</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>775.0</td>\n",
       "      <td>partner</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7986</th>\n",
       "      <td>paid</td>\n",
       "      <td>Retail</td>\n",
       "      <td>The Communal Bank &lt;em&gt;Las Caleras&lt;/em&gt; is loca...</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>275.0</td>\n",
       "      <td>partner</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7987</th>\n",
       "      <td>paid</td>\n",
       "      <td>Services</td>\n",
       "      <td>Eliud Ouma owns the Purple Hut salon located o...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>275.0</td>\n",
       "      <td>partner</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     status       sector                                                 en  \\\n",
       "7983   paid     Clothing  Rosa is a member of the 3 de Mayo Communal Ban...   \n",
       "7984   paid  Agriculture  The communal bank, \"Cumbe,\" is located on the ...   \n",
       "7985   paid         Food  Gladys Burgos Macas is a member of the 3 de Ma...   \n",
       "7986   paid       Retail  The Communal Bank <em>Las Caleras</em> is loca...   \n",
       "7987   paid     Services  Eliud Ouma owns the Purple Hut salon located o...   \n",
       "\n",
       "      country gender  loan_amount nonpayment  \n",
       "7983  Ecuador      F       1050.0     lender  \n",
       "7984  Ecuador      F       2650.0    partner  \n",
       "7985  Ecuador      F        775.0    partner  \n",
       "7986  Ecuador      F        275.0    partner  \n",
       "7987    Kenya      M        275.0    partner  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "in_dir = \"C:/Users/st50/Documents/sandbox/data\"\n",
    "out_dir = \"C:/Users/st50/Documents/sandbox/out\"\n",
    "\n",
    "%time kiva_df=pd.read_csv(os.path.join(in_dir, \"kiva.csv\"))\n",
    "#kiva_df = kiva_df.sample(frac=0.1)\n",
    "\n",
    "kiva_df.dropna(inplace=True)\n",
    "list(kiva_df)\n",
    "kiva_df.shape\n",
    "kiva_df.head()\n",
    "kiva_df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 4.59 s\n"
     ]
    },
    {
     "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>status</th>\n",
       "      <th>sector</th>\n",
       "      <th>en</th>\n",
       "      <th>country</th>\n",
       "      <th>gender</th>\n",
       "      <th>loan_amount</th>\n",
       "      <th>nonpayment</th>\n",
       "      <th>en_clean</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Agriculture</td>\n",
       "      <td>&lt;i&gt;This description is written by Rowland Amul...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>description written rowland amulyoto staff mem...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>&lt;i&gt;This description is written by Rowland Amul...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>description written rowland amulyoto staff mem...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>&lt;i&gt;This description was written by Richard Maz...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>description written richard mazengo staff memb...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Services</td>\n",
       "      <td>&lt;i&gt;This description was written by Rowland Amu...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>description written rowland amulyoto staff mem...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Construction</td>\n",
       "      <td>&lt;i&gt;This description was written by Rowland Amu...</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>description written rowland amulyoto staff mem...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      status        sector                                                 en  \\\n",
       "0  defaulted   Agriculture  <i>This description is written by Rowland Amul...   \n",
       "1  defaulted          Food  <i>This description is written by Rowland Amul...   \n",
       "2  defaulted          Food  <i>This description was written by Richard Maz...   \n",
       "3  defaulted      Services  <i>This description was written by Rowland Amu...   \n",
       "4  defaulted  Construction  <i>This description was written by Rowland Amu...   \n",
       "\n",
       "  country gender  loan_amount nonpayment  \\\n",
       "0   Kenya      M        500.0     lender   \n",
       "1   Kenya      F        500.0     lender   \n",
       "2   Kenya      M        500.0     lender   \n",
       "3   Kenya      F        500.0     lender   \n",
       "4   Kenya      M        500.0     lender   \n",
       "\n",
       "                                            en_clean  \n",
       "0  description written rowland amulyoto staff mem...  \n",
       "1  description written rowland amulyoto staff mem...  \n",
       "2  description written richard mazengo staff memb...  \n",
       "3  description written rowland amulyoto staff mem...  \n",
       "4  description written rowland amulyoto staff mem...  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from nltk.corpus import stopwords \n",
    "from bs4 import BeautifulSoup\n",
    "\n",
    "stop_words = set(stopwords.words('english') + stopwords.words('spanish'))\n",
    "\n",
    "lemmer = WordNetLemmatizer()\n",
    "\n",
    "def preprocess(x):\n",
    "    # Remove HTML tags\n",
    "    x = BeautifulSoup(x, \"lxml\").get_text()\n",
    "\n",
    "    # Lower case\n",
    "    x = x.lower()\n",
    "    \n",
    "    # Remove punctuation\n",
    "    x = re.sub(r'[^\\w\\s]', '', x)\n",
    "    \n",
    "    # Remove non-unicode\n",
    "    x = unidecode.unidecode(x)\n",
    "    \n",
    "    # Remove numbers\n",
    "    x = re.sub(r'\\d+', '', x)\n",
    "    \n",
    "    # Remove stopwords and lemmatize\n",
    "    x = [lemmer.lemmatize(w) for w in x.split() if w not in stop_words]\n",
    "    return ' '.join(x) \n",
    "\n",
    "%time kiva_df['en_clean'] = kiva_df['en'].apply(preprocess)\n",
    "kiva_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 4.01 s\n",
      "(7882, 342)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer\n",
    "\n",
    "no_features = 1000\n",
    "\n",
    "tf_vectorizer = CountVectorizer(min_df=.05, max_df=.5, max_features=no_features, ngram_range=[1,3])\n",
    "%time dtm_tf = tf_vectorizer.fit_transform(kiva_df['en_clean'])\n",
    "print(dtm_tf.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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>able</th>\n",
       "      <th>access</th>\n",
       "      <th>access business</th>\n",
       "      <th>access business training</th>\n",
       "      <th>active</th>\n",
       "      <th>active member</th>\n",
       "      <th>activity</th>\n",
       "      <th>add</th>\n",
       "      <th>additional</th>\n",
       "      <th>age</th>\n",
       "      <th>...</th>\n",
       "      <th>working</th>\n",
       "      <th>would</th>\n",
       "      <th>would like</th>\n",
       "      <th>wwwmifexorg</th>\n",
       "      <th>year ago</th>\n",
       "      <th>year experience</th>\n",
       "      <th>year old</th>\n",
       "      <th>year old married</th>\n",
       "      <th>yearold</th>\n",
       "      <th>young</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 342 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   able  access  access business  access business training  active  \\\n",
       "0     1       0                0                         0       0   \n",
       "1     0       0                0                         0       0   \n",
       "2     0       0                0                         0       0   \n",
       "3     0       0                0                         0       0   \n",
       "4     1       0                0                         0       0   \n",
       "\n",
       "   active member  activity  add  additional  age  ...    working  would  \\\n",
       "0              0         1    0           0    0  ...          0      0   \n",
       "1              0         0    0           0    0  ...          0      1   \n",
       "2              0         0    0           0    0  ...          0      0   \n",
       "3              0         0    0           0    0  ...          0      0   \n",
       "4              0         1    0           0    0  ...          0      0   \n",
       "\n",
       "   would like  wwwmifexorg  year ago  year experience  year old  \\\n",
       "0           0            0         0                0         0   \n",
       "1           0            0         0                0         0   \n",
       "2           0            0         0                0         0   \n",
       "3           0            0         0                0         0   \n",
       "4           0            0         0                0         0   \n",
       "\n",
       "   year old married  yearold  young  \n",
       "0                 0        0      0  \n",
       "1                 0        0      0  \n",
       "2                 0        0      0  \n",
       "3                 0        0      0  \n",
       "4                 0        0      0  \n",
       "\n",
       "[5 rows x 342 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(dtm_tf.toarray(), columns=tf_vectorizer.get_feature_names())\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Show the top tokens/grams"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('buy', 5359),\n",
       " ('also', 5169),\n",
       " ('school', 5148),\n",
       " ('group', 5086),\n",
       " ('family', 4768),\n",
       " ('able', 4550),\n",
       " ('work', 3951),\n",
       " ('old', 3929),\n",
       " ('one', 3811),\n",
       " ('small', 3780),\n",
       " ('product', 3728),\n",
       " ('help', 3712),\n",
       " ('two', 3673),\n",
       " ('year old', 3655),\n",
       " ('husband', 3488),\n",
       " ('member', 3264),\n",
       " ('community', 3212),\n",
       " ('woman', 3193),\n",
       " ('life', 3182),\n",
       " ('income', 3154)]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Calculate column sums from DTM\n",
    "sum_words = dtm_tf.sum(axis=0) \n",
    "words_freq = [(word, sum_words[0, idx]) for word, idx in tf_vectorizer.vocabulary_.items()]\n",
    "\n",
    "# Now, sort them\n",
    "words_freq =sorted(words_freq, key = lambda x: x[1], reverse=True)\n",
    "\n",
    "# Display top few\n",
    "words_freq[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2401eef7470>"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "FrequencyVisualizer(ax=<matplotlib.axes._subplots.AxesSubplot object at 0x00000240261C77B8>,\n",
       "          color=None,\n",
       "          features=['able', 'access', 'access business', 'access business training', 'active', 'active member', 'activity', 'add', 'additional', 'age', 'ago', 'allow', 'already', 'also', 'always', 'amount', 'another', 'applied', 'area', 'attend', 'back', 'bank', 'basic', 'bean', 'began', 'belief', 'better', '...', 'wwwmifexorg', 'year ago', 'year experience', 'year old', 'year old married', 'yearold', 'young'],\n",
       "          n=None, orient='h')"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2401eef7470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from yellowbrick.text import FreqDistVisualizer\n",
    "\n",
    "plt.figure(figsize=(5,8))\n",
    "visualizer = FreqDistVisualizer(features=tf_vectorizer.get_feature_names(), n=25)\n",
    "visualizer.fit(dtm_tf)\n",
    "visualizer.poof()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TSNEVisualizer(alpha=0.7,\n",
       "        ax=<matplotlib.axes._subplots.AxesSubplot object at 0x0000024021AE32E8>,\n",
       "        classes=None, colormap=None, colors=None, decompose=None,\n",
       "        decompose_by=None, labels=None, random_state=123)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x240261c7860>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from yellowbrick.text import TSNEVisualizer\n",
    "\n",
    "tsne = TSNEVisualizer(random_state=123)\n",
    "tsne.fit(dtm_tf, kiva_df['status'])\n",
    "tsne.poof()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Vectorization\n",
    "\n",
    "Using the cleaned Kiva data from above, we'll try a few different methods:\n",
    "\n",
    "- BOW (with raw TF, and TF-IDF)\n",
    "- Feature hashing\n",
    "- Topics\n",
    "- GloVe"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## BOW - TF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 3.98 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(7882, 342)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(7882, 348)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>status</th>\n",
       "      <th>sector</th>\n",
       "      <th>country</th>\n",
       "      <th>gender</th>\n",
       "      <th>loan_amount</th>\n",
       "      <th>nonpayment</th>\n",
       "      <th>able</th>\n",
       "      <th>access</th>\n",
       "      <th>access business</th>\n",
       "      <th>access business training</th>\n",
       "      <th>...</th>\n",
       "      <th>working</th>\n",
       "      <th>would</th>\n",
       "      <th>would like</th>\n",
       "      <th>wwwmifexorg</th>\n",
       "      <th>year ago</th>\n",
       "      <th>year experience</th>\n",
       "      <th>year old</th>\n",
       "      <th>year old married</th>\n",
       "      <th>yearold</th>\n",
       "      <th>young</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Agriculture</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Services</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Construction</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 348 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      status        sector country gender  loan_amount nonpayment  able  \\\n",
       "0  defaulted   Agriculture   Kenya      M        500.0     lender     1   \n",
       "1  defaulted          Food   Kenya      F        500.0     lender     0   \n",
       "2  defaulted          Food   Kenya      M        500.0     lender     0   \n",
       "3  defaulted      Services   Kenya      F        500.0     lender     0   \n",
       "4  defaulted  Construction   Kenya      M        500.0     lender     1   \n",
       "\n",
       "   access  access business  access business training  ...    working  would  \\\n",
       "0       0                0                         0  ...          0      0   \n",
       "1       0                0                         0  ...          0      1   \n",
       "2       0                0                         0  ...          0      0   \n",
       "3       0                0                         0  ...          0      0   \n",
       "4       0                0                         0  ...          0      0   \n",
       "\n",
       "   would like  wwwmifexorg  year ago  year experience  year old  \\\n",
       "0           0            0         0                0         0   \n",
       "1           0            0         0                0         0   \n",
       "2           0            0         0                0         0   \n",
       "3           0            0         0                0         0   \n",
       "4           0            0         0                0         0   \n",
       "\n",
       "   year old married  yearold  young  \n",
       "0                 0        0      0  \n",
       "1                 0        0      0  \n",
       "2                 0        0      0  \n",
       "3                 0        0      0  \n",
       "4                 0        0      0  \n",
       "\n",
       "[5 rows x 348 columns]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>status</th>\n",
       "      <th>sector</th>\n",
       "      <th>country</th>\n",
       "      <th>gender</th>\n",
       "      <th>loan_amount</th>\n",
       "      <th>nonpayment</th>\n",
       "      <th>able</th>\n",
       "      <th>access</th>\n",
       "      <th>access business</th>\n",
       "      <th>access business training</th>\n",
       "      <th>...</th>\n",
       "      <th>working</th>\n",
       "      <th>would</th>\n",
       "      <th>would like</th>\n",
       "      <th>wwwmifexorg</th>\n",
       "      <th>year ago</th>\n",
       "      <th>year experience</th>\n",
       "      <th>year old</th>\n",
       "      <th>year old married</th>\n",
       "      <th>yearold</th>\n",
       "      <th>young</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>7983</th>\n",
       "      <td>paid</td>\n",
       "      <td>Clothing</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>1050.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7984</th>\n",
       "      <td>paid</td>\n",
       "      <td>Agriculture</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>2650.0</td>\n",
       "      <td>partner</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7985</th>\n",
       "      <td>paid</td>\n",
       "      <td>Food</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>775.0</td>\n",
       "      <td>partner</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7986</th>\n",
       "      <td>paid</td>\n",
       "      <td>Retail</td>\n",
       "      <td>Ecuador</td>\n",
       "      <td>F</td>\n",
       "      <td>275.0</td>\n",
       "      <td>partner</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7987</th>\n",
       "      <td>paid</td>\n",
       "      <td>Services</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>275.0</td>\n",
       "      <td>partner</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 348 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     status       sector  country gender  loan_amount nonpayment  able  \\\n",
       "7983   paid     Clothing  Ecuador      F       1050.0     lender     0   \n",
       "7984   paid  Agriculture  Ecuador      F       2650.0    partner     2   \n",
       "7985   paid         Food  Ecuador      F        775.0    partner     0   \n",
       "7986   paid       Retail  Ecuador      F        275.0    partner     1   \n",
       "7987   paid     Services    Kenya      M        275.0    partner     0   \n",
       "\n",
       "      access  access business  access business training  ...    working  \\\n",
       "7983       0                0                         0  ...          0   \n",
       "7984       0                0                         0  ...          0   \n",
       "7985       0                0                         0  ...          0   \n",
       "7986       0                0                         0  ...          1   \n",
       "7987       0                0                         0  ...          0   \n",
       "\n",
       "      would  would like  wwwmifexorg  year ago  year experience  year old  \\\n",
       "7983      0           0            0         0                0         0   \n",
       "7984      0           0            0         0                0         0   \n",
       "7985      0           0            0         0                0         0   \n",
       "7986      0           0            0         0                0         1   \n",
       "7987      0           0            0         0                0         0   \n",
       "\n",
       "      year old married  yearold  young  \n",
       "7983                 0        0      0  \n",
       "7984                 0        4      0  \n",
       "7985                 0        0      0  \n",
       "7986                 0        0      0  \n",
       "7987                 0        1      0  \n",
       "\n",
       "[5 rows x 348 columns]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "\n",
    "tf_vectorizer = CountVectorizer(min_df=.05, max_df=.5, ngram_range=[1,3])\n",
    "%time dtm_tf = tf_vectorizer.fit_transform(kiva_df['en_clean'])\n",
    "\n",
    "bow_df = pd.DataFrame(dtm_tf.toarray(), columns=tf_vectorizer.get_feature_names(), index=kiva_df.index)\n",
    "bow_df.shape\n",
    "\n",
    "kiva_df_bow = pd.concat([kiva_df, bow_df], axis=1)\n",
    "kiva_df_bow.drop(columns=['en', 'en_clean'], inplace=True)\n",
    "kiva_df_bow.shape\n",
    "kiva_df_bow.head()\n",
    "kiva_df_bow.tail()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# BOW - TF-IDF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 4.1 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(7882, 342)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(7882, 348)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>status</th>\n",
       "      <th>sector</th>\n",
       "      <th>country</th>\n",
       "      <th>gender</th>\n",
       "      <th>loan_amount</th>\n",
       "      <th>nonpayment</th>\n",
       "      <th>able</th>\n",
       "      <th>access</th>\n",
       "      <th>access business</th>\n",
       "      <th>access business training</th>\n",
       "      <th>...</th>\n",
       "      <th>working</th>\n",
       "      <th>would</th>\n",
       "      <th>would like</th>\n",
       "      <th>wwwmifexorg</th>\n",
       "      <th>year ago</th>\n",
       "      <th>year experience</th>\n",
       "      <th>year old</th>\n",
       "      <th>year old married</th>\n",
       "      <th>yearold</th>\n",
       "      <th>young</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Agriculture</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.121480</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.106737</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Services</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Construction</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.105798</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 348 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      status        sector country gender  loan_amount nonpayment      able  \\\n",
       "0  defaulted   Agriculture   Kenya      M        500.0     lender  0.121480   \n",
       "1  defaulted          Food   Kenya      F        500.0     lender  0.000000   \n",
       "2  defaulted          Food   Kenya      M        500.0     lender  0.000000   \n",
       "3  defaulted      Services   Kenya      F        500.0     lender  0.000000   \n",
       "4  defaulted  Construction   Kenya      M        500.0     lender  0.105798   \n",
       "\n",
       "   access  access business  access business training  ...    working  \\\n",
       "0     0.0              0.0                       0.0  ...        0.0   \n",
       "1     0.0              0.0                       0.0  ...        0.0   \n",
       "2     0.0              0.0                       0.0  ...        0.0   \n",
       "3     0.0              0.0                       0.0  ...        0.0   \n",
       "4     0.0              0.0                       0.0  ...        0.0   \n",
       "\n",
       "      would  would like  wwwmifexorg  year ago  year experience  year old  \\\n",
       "0  0.000000         0.0          0.0       0.0              0.0       0.0   \n",
       "1  0.106737         0.0          0.0       0.0              0.0       0.0   \n",
       "2  0.000000         0.0          0.0       0.0              0.0       0.0   \n",
       "3  0.000000         0.0          0.0       0.0              0.0       0.0   \n",
       "4  0.000000         0.0          0.0       0.0              0.0       0.0   \n",
       "\n",
       "   year old married  yearold  young  \n",
       "0               0.0      0.0    0.0  \n",
       "1               0.0      0.0    0.0  \n",
       "2               0.0      0.0    0.0  \n",
       "3               0.0      0.0    0.0  \n",
       "4               0.0      0.0    0.0  \n",
       "\n",
       "[5 rows x 348 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "\n",
    "tfidf_vectorizer = TfidfVectorizer(min_df=.05, max_df=.5, ngram_range=[1,3])\n",
    "%time dtm_tfidf = tfidf_vectorizer.fit_transform(kiva_df['en_clean'])\n",
    "\n",
    "bow_df_tfidf = pd.DataFrame(dtm_tfidf.toarray(), columns=tfidf_vectorizer.get_feature_names(), index=kiva_df.index)\n",
    "bow_df_tfidf.shape\n",
    "\n",
    "kiva_df_bow_tfidf = pd.concat([kiva_df, bow_df_tfidf], axis=1)\n",
    "kiva_df_bow_tfidf.drop(columns=['en', 'en_clean'], inplace=True)\n",
    "kiva_df_bow_tfidf.shape\n",
    "kiva_df_bow_tfidf.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature Hashing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Simple Example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\st50\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\feature_extraction\\hashing.py:102: DeprecationWarning: the option non_negative=True has been deprecated in 0.19 and will be removed in version 0.21.\n",
      "  \" in version 0.21.\", DeprecationWarning)\n",
      "C:\\Users\\st50\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\feature_extraction\\hashing.py:102: DeprecationWarning: the option non_negative=True has been deprecated in 0.19 and will be removed in version 0.21.\n",
      "  \" in version 0.21.\", DeprecationWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[0., 0., 1., 0., 1., 2., 0., 1.],\n",
       "       [0., 0., 0., 1., 1., 1., 1., 1.],\n",
       "       [0., 1., 2., 0., 1., 3., 0., 0.]])"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(3, 8)"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     0    1    2    3    4    5    6    7\n",
       "0  0.0  0.0  1.0  0.0  1.0  2.0  0.0  1.0\n",
       "1  0.0  0.0  0.0  1.0  1.0  1.0  1.0  1.0\n",
       "2  0.0  1.0  2.0  0.0  1.0  3.0  0.0  0.0"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import HashingVectorizer\n",
    "\n",
    "docs = [\"my dog loves your dog\",\n",
    "        \"your dog loves the park\",\n",
    "        \"my dog and your dog love fleas\"]\n",
    "\n",
    "h_vectorizer = HashingVectorizer(n_features=2**3, ngram_range=[1,1], norm=None, non_negative=True, alternate_sign=False)\n",
    "dtm_h = h_vectorizer.fit_transform(docs)\n",
    "dtm_h.toarray()\n",
    "bow_df_h = pd.DataFrame(dtm_h.toarray())\n",
    "bow_df_h.shape\n",
    "bow_df_h.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Kiva Example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 1.14 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(7882, 512)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(7882, 518)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>status</th>\n",
       "      <th>sector</th>\n",
       "      <th>country</th>\n",
       "      <th>gender</th>\n",
       "      <th>loan_amount</th>\n",
       "      <th>nonpayment</th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>...</th>\n",
       "      <th>502</th>\n",
       "      <th>503</th>\n",
       "      <th>504</th>\n",
       "      <th>505</th>\n",
       "      <th>506</th>\n",
       "      <th>507</th>\n",
       "      <th>508</th>\n",
       "      <th>509</th>\n",
       "      <th>510</th>\n",
       "      <th>511</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Agriculture</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.063758</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.063758</td>\n",
       "      <td>0.127515</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.063758</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.048853</td>\n",
       "      <td>0.048853</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.048853</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.048853</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.048853</td>\n",
       "      <td>-0.048853</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Food</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.058926</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.058926</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Services</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>F</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.075593</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>defaulted</td>\n",
       "      <td>Construction</td>\n",
       "      <td>Kenya</td>\n",
       "      <td>M</td>\n",
       "      <td>500.0</td>\n",
       "      <td>lender</td>\n",
       "      <td>-0.067116</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.067116</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.067116</td>\n",
       "      <td>0.067116</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.067116</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 518 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      status        sector country gender  loan_amount nonpayment         0  \\\n",
       "0  defaulted   Agriculture   Kenya      M        500.0     lender  0.000000   \n",
       "1  defaulted          Food   Kenya      F        500.0     lender  0.048853   \n",
       "2  defaulted          Food   Kenya      M        500.0     lender  0.000000   \n",
       "3  defaulted      Services   Kenya      F        500.0     lender  0.000000   \n",
       "4  defaulted  Construction   Kenya      M        500.0     lender -0.067116   \n",
       "\n",
       "          1    2         3    ...     502  503  504       505       506  \\\n",
       "0  0.000000  0.0  0.000000    ...     0.0  0.0  0.0 -0.063758  0.000000   \n",
       "1  0.048853  0.0 -0.048853    ...     0.0  0.0  0.0  0.000000 -0.048853   \n",
       "2 -0.058926  0.0  0.000000    ...     0.0  0.0  0.0  0.000000  0.000000   \n",
       "3  0.000000  0.0  0.000000    ...     0.0  0.0  0.0  0.000000 -0.075593   \n",
       "4  0.000000  0.0  0.000000    ...     0.0  0.0  0.0  0.067116  0.000000   \n",
       "\n",
       "        507       508       509       510       511  \n",
       "0 -0.063758  0.127515  0.000000 -0.063758  0.000000  \n",
       "1  0.000000  0.000000 -0.048853 -0.048853  0.000000  \n",
       "2  0.058926  0.000000  0.000000  0.000000  0.000000  \n",
       "3  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "4  0.067116  0.067116  0.000000  0.000000  0.067116  \n",
       "\n",
       "[5 rows x 518 columns]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import HashingVectorizer\n",
    "\n",
    "h_vectorizer = HashingVectorizer(n_features=2**9, ngram_range=[1,3])\n",
    "%time dtm_h = h_vectorizer.fit_transform(kiva_df['en_clean'])\n",
    "bow_df_h = pd.DataFrame(dtm_h.toarray(), index=kiva_df.index)\n",
    "bow_df_h.shape\n",
    "\n",
    "kiva_df_h = pd.concat([kiva_df, bow_df_h], axis=1)\n",
    "kiva_df_h.drop(columns=['en', 'en_clean'], inplace=True)\n",
    "kiva_df_h.shape\n",
    "kiva_df_h.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Embeddings with Doc2Vec\n",
    "\n",
    "Inspirational blog posts:\n",
    "- https://www.analyticsvidhya.com/blog/2018/04/a-comprehensive-guide-to-understand-and-implement-text-classification-in-python/\n",
    "- https://markroxor.github.io/gensim/static/notebooks/doc2vec-lee.html"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Toy Model 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\st50\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\gensim\\utils.py:1197: UserWarning: detected Windows; aliasing chunkize to chunkize_serial\n",
      "  warnings.warn(\"detected Windows; aliasing chunkize to chunkize_serial\")\n"
     ]
    }
   ],
   "source": [
    "from gensim.models.doc2vec import Doc2Vec, TaggedDocument\n",
    "\n",
    "docs = [\"the dog is really smart\",\n",
    "        \"the cat is really smart\",\n",
    "        \"the rat is really smart\",\n",
    "        \"the mouse is really smart\",\n",
    "        \"the dog ran to the store\",\n",
    "        \"the cat ran to the store\",\n",
    "        \"the rat ran to the store\",\n",
    "        \"the mouse rand to the store\",\n",
    "        \"the dog is really dumb\",\n",
    "        \"the cat is really dumb\",\n",
    "        \"the dog is brown\",\n",
    "        \"the cat is brown\",\n",
    "        \"this document has nothing to do with any of the others\"]\n",
    "\n",
    "documents = [TaggedDocument(doc.split(), [i]) for i, doc in enumerate(docs)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 19.1 s\n",
      "Doc2Vec(dm/m,d5,n5,w2,s0.001,t4)\n",
      "[ 0.99248874  0.41061038  0.02131949  5.500582   -0.97168183]\n",
      "[-3.7883854 -1.4582045  1.6334202  0.6515057 -3.2550743]\n",
      "[-4.529103  -2.773157  -0.8444491  2.2315423 -1.759981 ]\n",
      "[-2.1608121  -1.8652055  -1.9994868   5.028724   -0.39297175]\n",
      "[0.9872789 3.5198617 2.4392228 1.7691325 4.3211737]\n",
      "[-3.2556016  1.585831   3.8298566 -2.5451846  2.4321797]\n",
      "[-4.310841   -0.42158404  1.0943148  -0.99977106  4.599478  ]\n",
      "[ 0.5099778  -1.716412   -0.13483465  3.6288497   6.089962  ]\n",
      "[-1.4100575  4.432139   0.5851983  4.7612267  1.3690388]\n",
      "[-5.8778386   2.5158525   1.8297108   0.31852654 -0.58893776]\n",
      "[ 2.2075603  3.1818073  3.8700628  2.948709  -1.1176034]\n",
      "[-1.439374    1.5633134   4.854304   -0.47629768 -2.7713668 ]\n",
      "[ 0.90785456 -5.2965612   4.4545765   2.5252194  -0.92053807]\n"
     ]
    }
   ],
   "source": [
    "%time model = Doc2Vec(documents, vector_size=5, window=2, min_count=1, workers=4, epochs=10000, alpha=0.065, seed=123)\n",
    "print(model)\n",
    "for i in range(0, len(docs)):\n",
    "    print(model.docvecs[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "Doc 0\n",
      "the dog is really smart\n",
      "[ 0.99248874  0.41061038  0.02131949  5.500582   -0.97168183]\n",
      "\n",
      "\n",
      "Closest doc 0 3 (0.7194470167160034)\n",
      "the mouse is really smart\n",
      "[-2.1608121  -1.8652055  -1.9994868   5.028724   -0.39297175]\n",
      "Closest doc 1 8 (0.65193772315979)\n",
      "the dog is really dumb\n",
      "[-1.4100575  4.432139   0.5851983  4.7612267  1.3690388]\n",
      "Closest doc 2 10 (0.5811523795127869)\n",
      "the dog is brown\n",
      "[ 2.2075603  3.1818073  3.8700628  2.948709  -1.1176034]\n",
      "\n",
      "\n",
      "Doc 1\n",
      "the cat is really smart\n",
      "[-3.7883854 -1.4582045  1.6334202  0.6515057 -3.2550743]\n",
      "\n",
      "\n",
      "Closest doc 0 2 (0.8083950877189636)\n",
      "the rat is really smart\n",
      "[-4.529103  -2.773157  -0.8444491  2.2315423 -1.759981 ]\n",
      "Closest doc 1 9 (0.6459209322929382)\n",
      "the cat is really dumb\n",
      "[-5.8778386   2.5158525   1.8297108   0.31852654 -0.58893776]\n",
      "Closest doc 2 11 (0.6013350486755371)\n",
      "the cat is brown\n",
      "[-1.439374    1.5633134   4.854304   -0.47629768 -2.7713668 ]\n",
      "\n",
      "\n",
      "Doc 2\n",
      "the rat is really smart\n",
      "[-4.529103  -2.773157  -0.8444491  2.2315423 -1.759981 ]\n",
      "\n",
      "\n",
      "Closest doc 0 1 (0.8083950877189636)\n",
      "the cat is really smart\n",
      "[-3.7883854 -1.4582045  1.6334202  0.6515057 -3.2550743]\n",
      "Closest doc 1 3 (0.7659169435501099)\n",
      "the mouse is really smart\n",
      "[-2.1608121  -1.8652055  -1.9994868   5.028724   -0.39297175]\n",
      "Closest doc 2 9 (0.48819655179977417)\n",
      "the cat is really dumb\n",
      "[-5.8778386   2.5158525   1.8297108   0.31852654 -0.58893776]\n",
      "\n",
      "\n",
      "Doc 3\n",
      "the mouse is really smart\n",
      "[-2.1608121  -1.8652055  -1.9994868   5.028724   -0.39297175]\n",
      "\n",
      "\n",
      "Closest doc 0 2 (0.7659169435501099)\n",
      "the rat is really smart\n",
      "[-4.529103  -2.773157  -0.8444491  2.2315423 -1.759981 ]\n",
      "Closest doc 1 0 (0.7194470167160034)\n",
      "the dog is really smart\n",
      "[ 0.99248874  0.41061038  0.02131949  5.500582   -0.97168183]\n",
      "Closest doc 2 8 (0.4069010019302368)\n",
      "the dog is really dumb\n",
      "[-1.4100575  4.432139   0.5851983  4.7612267  1.3690388]\n",
      "\n",
      "\n",
      "Doc 4\n",
      "the dog ran to the store\n",
      "[0.9872789 3.5198617 2.4392228 1.7691325 4.3211737]\n",
      "\n",
      "\n",
      "Closest doc 0 8 (0.6853932738304138)\n",
      "the dog is really dumb\n",
      "[-1.4100575  4.432139   0.5851983  4.7612267  1.3690388]\n",
      "Closest doc 1 7 (0.5729843378067017)\n",
      "the mouse rand to the store\n",
      "[ 0.5099778  -1.716412   -0.13483465  3.6288497   6.089962  ]\n",
      "Closest doc 2 10 (0.5728097558021545)\n",
      "the dog is brown\n",
      "[ 2.2075603  3.1818073  3.8700628  2.948709  -1.1176034]\n",
      "\n",
      "\n",
      "Doc 5\n",
      "the cat ran to the store\n",
      "[-3.2556016  1.585831   3.8298566 -2.5451846  2.4321797]\n",
      "\n",
      "\n",
      "Closest doc 0 6 (0.7606644034385681)\n",
      "the rat ran to the store\n",
      "[-4.310841   -0.42158404  1.0943148  -0.99977106  4.599478  ]\n",
      "Closest doc 1 9 (0.6583247184753418)\n",
      "the cat is really dumb\n",
      "[-5.8778386   2.5158525   1.8297108   0.31852654 -0.58893776]\n",
      "Closest doc 2 11 (0.5319925546646118)\n",
      "the cat is brown\n",
      "[-1.439374    1.5633134   4.854304   -0.47629768 -2.7713668 ]\n",
      "\n",
      "\n",
      "Doc 6\n",
      "the rat ran to the store\n",
      "[-4.310841   -0.42158404  1.0943148  -0.99977106  4.599478  ]\n",
      "\n",
      "\n",
      "Closest doc 0 5 (0.7606644034385681)\n",
      "the cat ran to the store\n",
      "[-3.2556016  1.585831   3.8298566 -2.5451846  2.4321797]\n",
      "Closest doc 1 9 (0.5360864996910095)\n",
      "the cat is really dumb\n",
      "[-5.8778386   2.5158525   1.8297108   0.31852654 -0.58893776]\n",
      "Closest doc 2 7 (0.4795892536640167)\n",
      "the mouse rand to the store\n",
      "[ 0.5099778  -1.716412   -0.13483465  3.6288497   6.089962  ]\n",
      "\n",
      "\n",
      "Doc 7\n",
      "the mouse rand to the store\n",
      "[ 0.5099778  -1.716412   -0.13483465  3.6288497   6.089962  ]\n",
      "\n",
      "\n",
      "Closest doc 0 4 (0.5729843378067017)\n",
      "the dog ran to the store\n",
      "[0.9872789 3.5198617 2.4392228 1.7691325 4.3211737]\n",
      "Closest doc 1 6 (0.4795892536640167)\n",
      "the rat ran to the store\n",
      "[-4.310841   -0.42158404  1.0943148  -0.99977106  4.599478  ]\n",
      "Closest doc 2 3 (0.4064691364765167)\n",
      "the mouse is really smart\n",
      "[-2.1608121  -1.8652055  -1.9994868   5.028724   -0.39297175]\n",
      "\n",
      "\n",
      "Doc 8\n",
      "the dog is really dumb\n",
      "[-1.4100575  4.432139   0.5851983  4.7612267  1.3690388]\n",
      "\n",
      "\n",
      "Closest doc 0 4 (0.6853932738304138)\n",
      "the dog ran to the store\n",
      "[0.9872789 3.5198617 2.4392228 1.7691325 4.3211737]\n",
      "Closest doc 1 0 (0.65193772315979)\n",
      "the dog is really smart\n",
      "[ 0.99248874  0.41061038  0.02131949  5.500582   -0.97168183]\n",
      "Closest doc 2 10 (0.5978721380233765)\n",
      "the dog is brown\n",
      "[ 2.2075603  3.1818073  3.8700628  2.948709  -1.1176034]\n",
      "\n",
      "\n",
      "Doc 9\n",
      "the cat is really dumb\n",
      "[-5.8778386   2.5158525   1.8297108   0.31852654 -0.58893776]\n",
      "\n",
      "\n",
      "Closest doc 0 5 (0.6583248376846313)\n",
      "the cat ran to the store\n",
      "[-3.2556016  1.585831   3.8298566 -2.5451846  2.4321797]\n",
      "Closest doc 1 1 (0.645920991897583)\n",
      "the cat is really smart\n",
      "[-3.7883854 -1.4582045  1.6334202  0.6515057 -3.2550743]\n",
      "Closest doc 2 11 (0.5675294995307922)\n",
      "the cat is brown\n",
      "[-1.439374    1.5633134   4.854304   -0.47629768 -2.7713668 ]\n",
      "\n",
      "\n",
      "Doc 10\n",
      "the dog is brown\n",
      "[ 2.2075603  3.1818073  3.8700628  2.948709  -1.1176034]\n",
      "\n",
      "\n",
      "Closest doc 0 8 (0.5978721976280212)\n",
      "the dog is really dumb\n",
      "[-1.4100575  4.432139   0.5851983  4.7612267  1.3690388]\n",
      "Closest doc 1 11 (0.5877246856689453)\n",
      "the cat is brown\n",
      "[-1.439374    1.5633134   4.854304   -0.47629768 -2.7713668 ]\n",
      "Closest doc 2 0 (0.5811524391174316)\n",
      "the dog is really smart\n",
      "[ 0.99248874  0.41061038  0.02131949  5.500582   -0.97168183]\n",
      "\n",
      "\n",
      "Doc 11\n",
      "the cat is brown\n",
      "[-1.439374    1.5633134   4.854304   -0.47629768 -2.7713668 ]\n",
      "\n",
      "\n",
      "Closest doc 0 1 (0.6013350486755371)\n",
      "the cat is really smart\n",
      "[-3.7883854 -1.4582045  1.6334202  0.6515057 -3.2550743]\n",
      "Closest doc 1 10 (0.5877246856689453)\n",
      "the dog is brown\n",
      "[ 2.2075603  3.1818073  3.8700628  2.948709  -1.1176034]\n",
      "Closest doc 2 9 (0.5675294995307922)\n",
      "the cat is really dumb\n",
      "[-5.8778386   2.5158525   1.8297108   0.31852654 -0.58893776]\n",
      "\n",
      "\n",
      "Doc 12\n",
      "this document has nothing to do with any of the others\n",
      "[ 0.90785456 -5.2965612   4.4545765   2.5252194  -0.92053807]\n",
      "\n",
      "\n",
      "Closest doc 0 1 (0.3943902254104614)\n",
      "the cat is really smart\n",
      "[-3.7883854 -1.4582045  1.6334202  0.6515057 -3.2550743]\n",
      "Closest doc 1 0 (0.3197977542877197)\n",
      "the dog is really smart\n",
      "[ 0.99248874  0.41061038  0.02131949  5.500582   -0.97168183]\n",
      "Closest doc 2 2 (0.309279203414917)\n",
      "the rat is really smart\n",
      "[-4.529103  -2.773157  -0.8444491  2.2315423 -1.759981 ]\n"
     ]
    }
   ],
   "source": [
    "for idx in range(0, len(docs)):\n",
    "    print(\"\\n\\nDoc {}\".format(idx))\n",
    "    print(docs[idx])\n",
    "    print(model.docvecs[idx])\n",
    "    sim = model.docvecs.most_similar(idx, topn=5)\n",
    "    print(\"\\n\")\n",
    "    \n",
    "    for i in range(0, 3):\n",
    "        print(\"Closest doc {} {} ({})\".format(i, sim[i][0], sim[i][1]))\n",
    "        closest_idx = sim[i][0]\n",
    "        print(docs[closest_idx])\n",
    "        print(model.docvecs[closest_idx])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Toy Model 3 - Kiva"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gensim.models.doc2vec import Doc2Vec, TaggedDocument\n",
    "\n",
    "docs = kiva_df['en_clean']\n",
    "\n",
    "documents = [TaggedDocument(doc.split(), [i]) for i, doc in enumerate(docs)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 2min 11s\n",
      "Doc2Vec(dm/m,d8,n5,w2,mc5,s0.001,t4)\n"
     ]
    }
   ],
   "source": [
    "%time model = Doc2Vec(documents, vector_size=8, window=2, min_count=5, workers=4, epochs=200, seed=123)\n",
    "print(model)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 2min 48s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "import collections\n",
    "\n",
    "ranks = []\n",
    "second_ranks = []\n",
    "for doc_id in range(len(documents)):\n",
    "    inferred_vector = model.infer_vector(documents[doc_id].words)\n",
    "    sims = model.docvecs.most_similar([inferred_vector], topn=len(model.docvecs))\n",
    "    rank = [docid for docid, sim in sims].index(doc_id)\n",
    "    ranks.append(rank)\n",
    "    second_ranks.append(sims[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(0, 467),\n",
       " (1, 313),\n",
       " (2, 256),\n",
       " (3, 231),\n",
       " (4, 204),\n",
       " (5, 191),\n",
       " (6, 167),\n",
       " (7, 151),\n",
       " (8, 146),\n",
       " (9, 139)]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "467"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "313"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "37"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "collections.Counter(ranks).most_common(10) \n",
    "collections.Counter(ranks)[0]\n",
    "collections.Counter(ranks)[1]\n",
    "collections.Counter(ranks)[43]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Document (4554): «bolvar year experience manufacturing dresser storing clothes left native city came live near uncle isla trinitaira guayaquil started work learned profession slowly became independent two year ago married wife charge buying wood business located front place live dont pay rent caretaker place steady client order product every week charge advance balance upon delivery work monday saturday trying little little get ahead still paying used planing machine bought bolvar need loan buy another planing machine always employee working right need take turn permit production increase bolvar year old one child wife help business especially buying wood married two year since young trying get ahead family life well business aspiration establish business well save buy home translated spanish naomi johnson kiva volunteer»\n",
      "\n",
      "SIMILAR/DISSIMILAR DOCS PER MODEL Doc2Vec(dm/m,d8,n5,w2,mc5,s0.001,t4):\n",
      "\n",
      "MOST (6029, 0.9597870111465454): «deanis altagracia guzman wish build full scale beauty center featuring acrylic nail fixing capacity deanis living year husband three kid recently started selling clothing well since need complement income produced beauty aid service product also plan sell clothing family usa sends time time husband finish construction area plan move salon operation plan able assist helped feel lucky grateful able undertake thus far»\n",
      "\n",
      "MEDIAN (1234, 0.4166688919067383): «jane njoki year old married child jane never went school due effort educated child jane active member namelok woman group member baptist church kipayu started business steer help weec loan sale steer grazing period managed buy steer due determination steer profit earns enables repay group saving educate child jane want loan usd buy steer business»\n",
      "\n",
      "LEAST (5739, -0.2855098843574524): «young rolando farming past year day gaining experience cultivation rice profitable crop region rolando want use money purchase farming implement order productive short growing season currently rolando single child life house made reed live past year dream one day land farm work improving home making stable plan lot effort dedication translated spanish carolina zelayakronlund kiva volunteer»\n",
      "\n"
     ]
    }
   ],
   "source": [
    "doc_id = 4554\n",
    "print('Document ({}): «{}»\\n'.format(doc_id, ' '.join(documents[doc_id].words)))\n",
    "print(u'SIMILAR/DISSIMILAR DOCS PER MODEL %s:\\n' % model)\n",
    "for label, index in [('MOST', 0), ('MEDIAN', len(sims)//2), ('LEAST', len(sims) - 1)]:\n",
    "    print(u'%s %s: «%s»\\n' % (label, sims[index], ' '.join(documents[sims[index][0]].words)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Document (543): «mary year old married child active member st jude woman group located githunguri thika unit christian faith member presbyterian church east africa teacher local school also practice farming rear dairy cattle grows vegetable farm earns total usd per month teaching job farming plan make sure child get good education provide best mother give biggest challenge finance feel improve farming method better position usd loan would like buy organic manure soil supplement increase production farm willing make repayment monthly basis»\n",
      "\n",
      "Similar Document (396, 0.9827365279197693): «mary gakenia maina year old single mother two child one child junior school second year old active member wendani self help group life ngorika settlement rift valley part kenya one acre piece land farm maize potato potato farming seasonal done four time year give competitive price planted right time given appropriate attention term water pest control requesting loan use follows lease two acre land buy potato seedling maize seed fertilizer buy pesticide sprayer pesticide plan save net profit yield offer better planned future two child since breadwinner»\n",
      "\n"
     ]
    }
   ],
   "source": [
    "doc_id = 543\n",
    "print('Train Document ({}): «{}»\\n'.format(doc_id, ' '.join(documents[doc_id].words)))\n",
    "sim_id = second_ranks[doc_id]\n",
    "print('Similar Document {}: «{}»\\n'.format(sim_id, ' '.join(documents[sim_id[0]].words)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(1712, 0.9852263927459717),\n",
       " (1170, 0.9747959971427917),\n",
       " (432, 0.9694892168045044),\n",
       " (2923, 0.9689468741416931),\n",
       " (3576, 0.9681254029273987),\n",
       " (954, 0.9673334360122681),\n",
       " (4276, 0.9670917987823486),\n",
       " (4277, 0.9649144411087036),\n",
       " (3429, 0.9647232294082642),\n",
       " (814, 0.9634026885032654)]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "'description written rowland amulyoto staff member village enterprise fund representing business robert married child addition family take care mother brother robert started planting vegetable selling local market diversified tea nursery profitable given loan robert able improve activity buying fertilizer pesticide pump seedling pump remove uncertainty weather current capital venture ksh allow robert maximize potential robert completed high school never got employment got apprentice training tea extension officer go getter main hobby teaching music'"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "array([ 1.1980051 ,  1.023994  , -3.0658832 , -1.3409452 ,  1.1148725 ,\n",
       "       -2.077089  ,  0.19002658, -0.19003806], dtype=float32)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Closest doc 0 1712 (0.9852263927459717)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'teresia wanjiku born year ago town kiambu married four child first born working hotel nairobi two college fourth still primary school dedicated christian worship kanunga catholic church member catholic woman association cwa teresias obvious leadership skill made role model many young woman year since marriage husband run tailoring shop local shopping center working dairy farmer also raise pig earlier year friend told teresia many way woman economic empowerment consort kiva working boost business kiambu became interested project immediately joined woman group come u loan use boost tailoring business buying clothing material use make dress intends repay loan within month'"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "array([ 1.3800128 ,  1.0005883 , -3.229257  , -1.3149457 ,  1.5749676 ,\n",
       "       -1.5221534 ,  0.13650416, -0.17150591], dtype=float32)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Closest doc 1 1170 (0.9747959971427917)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'susan born st march married one child lower primary school private primary school nairobi susan like colleague joined group registered weec hearing good news weec kiva boosted many business kiambu surroundings elected discipline mistress group susan sell secondhand clothes kawangware business last year raised educated child income earned secondhand clothes business given loan usd intends buy bale secondhand clothes willing repay loan within month'"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "array([ 1.9684074 ,  2.007368  , -5.332796  , -2.4643576 ,  1.7331749 ,\n",
       "       -2.0837624 , -0.00788527, -0.99241424], dtype=float32)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Closest doc 2 432 (0.9694892168045044)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'jane married five child child attend public primary school isinya kenya active member also treasurer nanapae woman group also treasurer merry go round area emily joined group registered weec year jane started bead work business four year ago initial capital amount usd loaned weec make beautiful belt mat dress shoe earring conduct business open air market isinya stock increased usd usd jane seeking loan usd add additional stock business'"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "array([ 1.9800103 ,  2.120033  , -4.589822  , -1.7324752 ,  1.300927  ,\n",
       "       -1.6647959 ,  0.33758453, -0.24634889], dtype=float32)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "idx = 0\n",
    "sim = model.docvecs.most_similar(idx, topn=10)\n",
    "sim\n",
    "\n",
    "idx\n",
    "docs.iloc[idx]\n",
    "model.docvecs[idx]\n",
    "\n",
    "for i in range(0, 3):\n",
    "    print(\"Closest doc {} {} ({})\".format(i, sim[i][0], sim[i][1]))\n",
    "    closest_idx = sim[i][0]\n",
    "    docs.iloc[closest_idx]\n",
    "    model.docvecs[closest_idx]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'saami year old married child religiuou together wither family saami primary education class married completing primary level education due cultural limitationssaami treasurer kiret naboisho w g attended various leadership training session organized weec training assist woman running group activity saami steer however died drought period death almost whole stock affected family income generating activity saami plan build stock intends buy steer usd sell steer month reinvest proceeds steer business saami intends attend livestock management course scheduled december saami feel important train modern method managing livestock also listed train basic business management skill'"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "docs.iloc[25]"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}