notebook/IML-Project-Final.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Project Work - Sentimental Analysis on Amazon Review(Music Instruments)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "dataset - http://jmcauley.ucsd.edu/data/amazon/\n",
    "\n",
    "#### Musical Instruments  Prodcut Review is used. \n",
    "\n",
    "### Experiment Setup Overview\n",
    "- Derived a small Dataset from 10 Reviews which has all kind of reviews(5, 4, 3, 2, 1 stars)\n",
    "- Created a seperate training and test dataset. Test is not used in any training.\n",
    "- Human Labelled from My friend <b><span style=\"color:purple\">Hemanth and Bashyeam. - Thank you Guys</span></b>  .50 Documents are carefully labelled. \n",
    "- Everything is exported into csv. That is used in this experiment.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd \n",
    "import json\n",
    "import pymongo\n",
    "from pymongo import MongoClient\n",
    "from bson.objectid import ObjectId \n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.naive_bayes import BernoulliNB\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn import svm\n",
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn import tree\n",
    "import graphviz\n",
    "import sklearn.metrics\n",
    "from sklearn import metrics\n",
    "import re\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.feature_selection import chi2\n",
    "import pymongo\n",
    "import random\n",
    "from pymongo import MongoClient\n",
    "pd.set_option('display.max_columns', None)\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 1: Dataset Description"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Amazon review dataset on Musical Instuments"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Train-Columns</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>_id</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>asin</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>cleaned_review_text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>helpful.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>helpful.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>human_labelled</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>keywords</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>labelled</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>load_to_db</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>neg_doc</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>overall</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>rationales</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>rationals.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>rationals.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>rationals.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>rationals.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>rationals.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>rationals.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>rationals.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>rationals.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>reviewText</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Train-Columns\n",
       "0                   _id\n",
       "1                  asin\n",
       "2   cleaned_review_text\n",
       "3             helpful.0\n",
       "4             helpful.1\n",
       "5        human_labelled\n",
       "6              keywords\n",
       "7              labelled\n",
       "8            load_to_db\n",
       "9               neg_doc\n",
       "10              overall\n",
       "11           rationales\n",
       "12          rationals.0\n",
       "13          rationals.1\n",
       "14          rationals.2\n",
       "15          rationals.3\n",
       "16          rationals.4\n",
       "17          rationals.5\n",
       "18          rationals.6\n",
       "19          rationals.7\n",
       "20           reviewText"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train = pd.read_csv(\"../data/dataset.csv\",header=0)\n",
    "test = pd.read_csv(\"../data/test.csv\",header=0)\n",
    "train = train.fillna('')\n",
    "pd.DataFrame(train.columns, columns=['Train-Columns'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Test-Columns</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>asin</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>overall</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>reviewText</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Test-Columns\n",
       "0         asin\n",
       "1      overall\n",
       "2   reviewText"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(test.columns, columns=['Test-Columns'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>asin</th>\n",
       "      <th>overall</th>\n",
       "      <th>reviewText</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>B0002E37MM</td>\n",
       "      <td>1</td>\n",
       "      <td>I bought this hoping to use as intended, angle...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>B0002E2Z8Y</td>\n",
       "      <td>1</td>\n",
       "      <td>I know many people love the design, but I find...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>B0002E2KPC</td>\n",
       "      <td>1</td>\n",
       "      <td>This is an update to my original review becaus...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>B0002E1O2M</td>\n",
       "      <td>1</td>\n",
       "      <td>I bought a set as they were recommended to get...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>B0002DV7U2</td>\n",
       "      <td>1</td>\n",
       "      <td>...for something, but not a guitar.  Wasn't su...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         asin  overall                                         reviewText\n",
       "0  B0002E37MM        1  I bought this hoping to use as intended, angle...\n",
       "1  B0002E2Z8Y        1  I know many people love the design, but I find...\n",
       "2  B0002E2KPC        1  This is an update to my original review becaus...\n",
       "3  B0002E1O2M        1  I bought a set as they were recommended to get...\n",
       "4  B0002DV7U2        1  ...for something, but not a guitar.  Wasn't su..."
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>_id</th>\n",
       "      <th>asin</th>\n",
       "      <th>cleaned_review_text</th>\n",
       "      <th>helpful.0</th>\n",
       "      <th>helpful.1</th>\n",
       "      <th>human_labelled</th>\n",
       "      <th>keywords</th>\n",
       "      <th>labelled</th>\n",
       "      <th>load_to_db</th>\n",
       "      <th>neg_doc</th>\n",
       "      <th>overall</th>\n",
       "      <th>rationales</th>\n",
       "      <th>rationals.0</th>\n",
       "      <th>rationals.1</th>\n",
       "      <th>rationals.2</th>\n",
       "      <th>rationals.3</th>\n",
       "      <th>rationals.4</th>\n",
       "      <th>rationals.5</th>\n",
       "      <th>rationals.6</th>\n",
       "      <th>rationals.7</th>\n",
       "      <th>reviewText</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5a1e177778b95010ea1895d2</td>\n",
       "      <td>1384719342</td>\n",
       "      <td>Not much to write about here but it does exact...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td></td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td></td>\n",
       "      <td>5.0</td>\n",
       "      <td></td>\n",
       "      <td>exactly</td>\n",
       "      <td>crisp</td>\n",
       "      <td>lowest</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>Not much to write about here, but it does exac...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5a1e177778b95010ea1895d3</td>\n",
       "      <td>1384719342</td>\n",
       "      <td>The product does exactly as it should and is q...</td>\n",
       "      <td>13</td>\n",
       "      <td>14</td>\n",
       "      <td>False</td>\n",
       "      <td></td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td></td>\n",
       "      <td>5.0</td>\n",
       "      <td></td>\n",
       "      <td>does</td>\n",
       "      <td>affordable</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>The product does exactly as it should and is q...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>5a1e177778b95010ea1895d4</td>\n",
       "      <td>1384719342</td>\n",
       "      <td>The primary job of this device is to block the...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "      <td></td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td></td>\n",
       "      <td>5.0</td>\n",
       "      <td></td>\n",
       "      <td>no</td>\n",
       "      <td>noticeable</td>\n",
       "      <td>reduction</td>\n",
       "      <td>secure</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>The primary job of this device is to block the...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        _id        asin  \\\n",
       "0  5a1e177778b95010ea1895d2  1384719342   \n",
       "1  5a1e177778b95010ea1895d3  1384719342   \n",
       "2  5a1e177778b95010ea1895d4  1384719342   \n",
       "\n",
       "                                 cleaned_review_text  helpful.0  helpful.1  \\\n",
       "0  Not much to write about here but it does exact...          0          0   \n",
       "1  The product does exactly as it should and is q...         13         14   \n",
       "2  The primary job of this device is to block the...          1          1   \n",
       "\n",
       "   human_labelled keywords  labelled  load_to_db neg_doc  overall rationales  \\\n",
       "0           False               True        True              5.0              \n",
       "1           False               True        True              5.0              \n",
       "2           False               True        True              5.0              \n",
       "\n",
       "  rationals.0 rationals.1 rationals.2 rationals.3 rationals.4 rationals.5  \\\n",
       "0     exactly       crisp      lowest                                       \n",
       "1        does  affordable                                                   \n",
       "2          no  noticeable   reduction      secure                           \n",
       "\n",
       "  rationals.6 rationals.7                                         reviewText  \n",
       "0                          Not much to write about here, but it does exac...  \n",
       "1                          The product does exactly as it should and is q...  \n",
       "2                          The primary job of this device is to block the...  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Section 2: Model Fit Basic"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Logistic Regression and BNB is used\n",
    "For testing what is accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = train['cleaned_review_text']\n",
    "X_test = test['reviewText']\n",
    "train['Y'] = train['overall'] == 5.0\n",
    "test['Y'] = test['overall'] == 5.0\n",
    "train['Y'] = train['overall'] == 4.0\n",
    "test['Y'] = test['overall'] == 4.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_train = []\n",
    "y_test = []\n",
    "for each in range(train.shape[0]):\n",
    "    if train.loc[each, 'Y']:\n",
    "        y_train.append(1)\n",
    "        y_test.append(1)\n",
    "    else:\n",
    "        y_train.append(0)\n",
    "        y_test.append(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Positve Docs 343, Negative Docs 183'"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f\"Positve Docs {np.sum(y_train)}, Negative Docs {train.shape[0] - np.sum(y_train)}\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "vectorizer = CountVectorizer(min_df=5, max_df=1.0, binary=True)\n",
    "X_train_vec = vectorizer.fit_transform(X_train)\n",
    "X_test_vec = vectorizer.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "features = pd.DataFrame(vectorizer.get_feature_names())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BernoulliNB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.81939163498098855"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = BernoulliNB()\n",
    "clf.fit(X_train_vec, y_train)\n",
    "\n",
    "y_predict = clf.predict(X_train_vec)\n",
    "y_proba = clf.predict_proba(X_test_vec)\n",
    "\n",
    "accuracy_score(y_test, y_predict)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### AUC Score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11222c588>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# calculate the fpr and tpr for all thresholds of the classification\n",
    "probs = clf.predict_proba(X_train_vec)\n",
    "preds = probs[:,1]\n",
    "fpr, tpr, threshold = metrics.roc_curve(y_test, preds)\n",
    "roc_auc = metrics.auc(fpr, tpr)\n",
    "\n",
    "# method I: plt\n",
    "import matplotlib.pyplot as plt\n",
    "plt.title('Receiver Operating Characteristic')\n",
    "plt.plot(fpr, tpr, 'b', label = 'AUC = %0.2f' % roc_auc)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.plot([0, 1], [0, 1],'r--')\n",
    "plt.xlim([0, 1])\n",
    "plt.ylim([0, 1])\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Logistic Regression"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### C = 0.001  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.65209125475285168"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = LogisticRegression(penalty='l2', C = 0.001)\n",
    "clf.fit(X_train_vec, y_train)\n",
    "y_predict = clf.predict(X_train_vec)\n",
    "y_proba = clf.predict_proba(X_test_vec)\n",
    "\n",
    "accuracy_score(y_test, y_predict)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### AUC Score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMcaYpCxRGGOMSer/ATKR/JaCinwuAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10bed7e48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# calculate the fpr and tpr for all thresholds of the classification\n",
    "probs = clf.predict_proba(X_train_vec)\n",
    "preds = probs[:,1]\n",
    "fpr, tpr, threshold = metrics.roc_curve(y_test, preds)\n",
    "roc_auc = metrics.auc(fpr, tpr)\n",
    "\n",
    "# method I: plt\n",
    "import matplotlib.pyplot as plt\n",
    "plt.title('Receiver Operating Characteristic')\n",
    "plt.plot(fpr, tpr, 'b', label = 'AUC = %0.2f' % roc_auc)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.plot([0, 1], [0, 1],'r--')\n",
    "plt.xlim([0, 1])\n",
    "plt.ylim([0, 1])\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### C = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.99619771863117867"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = LogisticRegression(penalty='l2', C = 1)\n",
    "clf.fit(X_train_vec, y_train)\n",
    "y_predict = clf.predict(X_train_vec)\n",
    "y_proba = clf.predict_proba(X_test_vec)\n",
    "\n",
    "accuracy_score(y_test, y_predict)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### AUC Score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11234b710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# calculate the fpr and tpr for all thresholds of the classification\n",
    "probs = clf.predict_proba(X_train_vec)\n",
    "preds = probs[:,1]\n",
    "fpr, tpr, threshold = metrics.roc_curve(y_test, preds)\n",
    "roc_auc = metrics.auc(fpr, tpr)\n",
    "\n",
    "# method I: plt\n",
    "import matplotlib.pyplot as plt\n",
    "plt.title('Receiver Operating Characteristic')\n",
    "plt.plot(fpr, tpr, 'b', label = 'AUC = %0.2f' % roc_auc)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.plot([0, 1], [0, 1],'r--')\n",
    "plt.xlim([0, 1])\n",
    "plt.ylim([0, 1])\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "'''\n",
    "step 1: It converts the co-ef to absolute value\n",
    "step 2: Appends to info list(feature_name, weight)\n",
    "'''\n",
    "def top_weighted_features(coef,label, features):\n",
    "    coef = np.ravel(np.abs(coef))\n",
    "    info = []\n",
    "    index = np.ravel(coef).argsort()[::-1][:11]\n",
    "    for t in range(10):\n",
    "        info.append([features[index[t]][0],coef[index[t]]])\n",
    "    return pd.DataFrame(info)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "features = pd.DataFrame.transpose(features)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Top 10 Feature absolute weights "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Transparency "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>midi</td>\n",
       "      <td>1.189383</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>fine</td>\n",
       "      <td>1.072322</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>noticed</td>\n",
       "      <td>1.045061</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>came</td>\n",
       "      <td>0.970896</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>thick</td>\n",
       "      <td>0.925214</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>just</td>\n",
       "      <td>0.923685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>almost</td>\n",
       "      <td>0.870380</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>equipment</td>\n",
       "      <td>0.860073</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>noisy</td>\n",
       "      <td>0.859717</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>by</td>\n",
       "      <td>0.837260</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           0         1\n",
       "0       midi  1.189383\n",
       "1       fine  1.072322\n",
       "2    noticed  1.045061\n",
       "3       came  0.970896\n",
       "4      thick  0.925214\n",
       "5       just  0.923685\n",
       "6     almost  0.870380\n",
       "7  equipment  0.860073\n",
       "8      noisy  0.859717\n",
       "9         by  0.837260"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top_weighted_features(clf.coef_[0],'LR', features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\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",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>10</th>\n",
       "      <th>11</th>\n",
       "      <th>12</th>\n",
       "      <th>13</th>\n",
       "      <th>14</th>\n",
       "      <th>15</th>\n",
       "      <th>16</th>\n",
       "      <th>17</th>\n",
       "      <th>18</th>\n",
       "      <th>19</th>\n",
       "      <th>20</th>\n",
       "      <th>21</th>\n",
       "      <th>22</th>\n",
       "      <th>23</th>\n",
       "      <th>24</th>\n",
       "      <th>25</th>\n",
       "      <th>26</th>\n",
       "      <th>27</th>\n",
       "      <th>28</th>\n",
       "      <th>29</th>\n",
       "      <th>30</th>\n",
       "      <th>31</th>\n",
       "      <th>32</th>\n",
       "      <th>33</th>\n",
       "      <th>34</th>\n",
       "      <th>35</th>\n",
       "      <th>36</th>\n",
       "      <th>37</th>\n",
       "      <th>38</th>\n",
       "      <th>39</th>\n",
       "      <th>40</th>\n",
       "      <th>41</th>\n",
       "      <th>42</th>\n",
       "      <th>43</th>\n",
       "      <th>44</th>\n",
       "      <th>45</th>\n",
       "      <th>46</th>\n",
       "      <th>47</th>\n",
       "      <th>48</th>\n",
       "      <th>49</th>\n",
       "      <th>50</th>\n",
       "      <th>51</th>\n",
       "      <th>52</th>\n",
       "      <th>53</th>\n",
       "      <th>54</th>\n",
       "      <th>55</th>\n",
       "      <th>56</th>\n",
       "      <th>57</th>\n",
       "      <th>58</th>\n",
       "      <th>59</th>\n",
       "      <th>60</th>\n",
       "      <th>61</th>\n",
       "      <th>62</th>\n",
       "      <th>63</th>\n",
       "      <th>64</th>\n",
       "      <th>65</th>\n",
       "      <th>66</th>\n",
       "      <th>67</th>\n",
       "      <th>68</th>\n",
       "      <th>69</th>\n",
       "      <th>70</th>\n",
       "      <th>71</th>\n",
       "      <th>72</th>\n",
       "      <th>73</th>\n",
       "      <th>74</th>\n",
       "      <th>75</th>\n",
       "      <th>76</th>\n",
       "      <th>77</th>\n",
       "      <th>78</th>\n",
       "      <th>79</th>\n",
       "      <th>80</th>\n",
       "      <th>81</th>\n",
       "      <th>82</th>\n",
       "      <th>83</th>\n",
       "      <th>84</th>\n",
       "      <th>85</th>\n",
       "      <th>86</th>\n",
       "      <th>87</th>\n",
       "      <th>88</th>\n",
       "      <th>89</th>\n",
       "      <th>90</th>\n",
       "      <th>91</th>\n",
       "      <th>92</th>\n",
       "      <th>93</th>\n",
       "      <th>94</th>\n",
       "      <th>95</th>\n",
       "      <th>96</th>\n",
       "      <th>97</th>\n",
       "      <th>98</th>\n",
       "      <th>99</th>\n",
       "      <th>100</th>\n",
       "      <th>101</th>\n",
       "      <th>102</th>\n",
       "      <th>103</th>\n",
       "      <th>104</th>\n",
       "      <th>105</th>\n",
       "      <th>106</th>\n",
       "      <th>107</th>\n",
       "      <th>108</th>\n",
       "      <th>109</th>\n",
       "      <th>110</th>\n",
       "      <th>111</th>\n",
       "      <th>112</th>\n",
       "      <th>113</th>\n",
       "      <th>114</th>\n",
       "      <th>115</th>\n",
       "      <th>116</th>\n",
       "      <th>117</th>\n",
       "      <th>118</th>\n",
       "      <th>119</th>\n",
       "      <th>120</th>\n",
       "      <th>121</th>\n",
       "      <th>122</th>\n",
       "      <th>123</th>\n",
       "      <th>124</th>\n",
       "      <th>125</th>\n",
       "      <th>126</th>\n",
       "      <th>127</th>\n",
       "      <th>128</th>\n",
       "      <th>129</th>\n",
       "      <th>130</th>\n",
       "      <th>131</th>\n",
       "      <th>132</th>\n",
       "      <th>133</th>\n",
       "      <th>134</th>\n",
       "      <th>135</th>\n",
       "      <th>136</th>\n",
       "      <th>137</th>\n",
       "      <th>138</th>\n",
       "      <th>139</th>\n",
       "      <th>140</th>\n",
       "      <th>141</th>\n",
       "      <th>142</th>\n",
       "      <th>143</th>\n",
       "      <th>144</th>\n",
       "      <th>145</th>\n",
       "      <th>146</th>\n",
       "      <th>147</th>\n",
       "      <th>148</th>\n",
       "      <th>149</th>\n",
       "      <th>150</th>\n",
       "      <th>151</th>\n",
       "      <th>152</th>\n",
       "      <th>153</th>\n",
       "      <th>154</th>\n",
       "      <th>155</th>\n",
       "      <th>156</th>\n",
       "      <th>157</th>\n",
       "      <th>158</th>\n",
       "      <th>159</th>\n",
       "      <th>160</th>\n",
       "      <th>161</th>\n",
       "      <th>162</th>\n",
       "      <th>163</th>\n",
       "      <th>164</th>\n",
       "      <th>165</th>\n",
       "      <th>166</th>\n",
       "      <th>167</th>\n",
       "      <th>168</th>\n",
       "      <th>169</th>\n",
       "      <th>170</th>\n",
       "      <th>171</th>\n",
       "      <th>172</th>\n",
       "      <th>173</th>\n",
       "      <th>174</th>\n",
       "      <th>175</th>\n",
       "      <th>176</th>\n",
       "      <th>177</th>\n",
       "      <th>178</th>\n",
       "      <th>179</th>\n",
       "      <th>180</th>\n",
       "      <th>181</th>\n",
       "      <th>182</th>\n",
       "      <th>183</th>\n",
       "      <th>184</th>\n",
       "      <th>185</th>\n",
       "      <th>186</th>\n",
       "      <th>187</th>\n",
       "      <th>188</th>\n",
       "      <th>189</th>\n",
       "      <th>190</th>\n",
       "      <th>191</th>\n",
       "      <th>192</th>\n",
       "      <th>193</th>\n",
       "      <th>194</th>\n",
       "      <th>195</th>\n",
       "      <th>196</th>\n",
       "      <th>197</th>\n",
       "      <th>198</th>\n",
       "      <th>199</th>\n",
       "      <th>200</th>\n",
       "      <th>201</th>\n",
       "      <th>202</th>\n",
       "      <th>203</th>\n",
       "      <th>204</th>\n",
       "      <th>205</th>\n",
       "      <th>206</th>\n",
       "      <th>207</th>\n",
       "      <th>208</th>\n",
       "      <th>209</th>\n",
       "      <th>210</th>\n",
       "      <th>211</th>\n",
       "      <th>212</th>\n",
       "      <th>213</th>\n",
       "      <th>214</th>\n",
       "      <th>215</th>\n",
       "      <th>216</th>\n",
       "      <th>217</th>\n",
       "      <th>218</th>\n",
       "      <th>219</th>\n",
       "      <th>220</th>\n",
       "      <th>221</th>\n",
       "      <th>222</th>\n",
       "      <th>223</th>\n",
       "      <th>224</th>\n",
       "      <th>225</th>\n",
       "      <th>226</th>\n",
       "      <th>227</th>\n",
       "      <th>228</th>\n",
       "      <th>229</th>\n",
       "      <th>230</th>\n",
       "      <th>231</th>\n",
       "      <th>232</th>\n",
       "      <th>233</th>\n",
       "      <th>234</th>\n",
       "      <th>235</th>\n",
       "      <th>236</th>\n",
       "      <th>237</th>\n",
       "      <th>238</th>\n",
       "      <th>239</th>\n",
       "      <th>240</th>\n",
       "      <th>241</th>\n",
       "      <th>242</th>\n",
       "      <th>243</th>\n",
       "      <th>244</th>\n",
       "      <th>245</th>\n",
       "      <th>246</th>\n",
       "      <th>247</th>\n",
       "      <th>248</th>\n",
       "      <th>249</th>\n",
       "      <th>250</th>\n",
       "      <th>251</th>\n",
       "      <th>252</th>\n",
       "      <th>253</th>\n",
       "      <th>254</th>\n",
       "      <th>255</th>\n",
       "      <th>256</th>\n",
       "      <th>257</th>\n",
       "      <th>258</th>\n",
       "      <th>259</th>\n",
       "      <th>260</th>\n",
       "      <th>261</th>\n",
       "      <th>262</th>\n",
       "      <th>263</th>\n",
       "      <th>264</th>\n",
       "      <th>265</th>\n",
       "      <th>266</th>\n",
       "      <th>267</th>\n",
       "      <th>268</th>\n",
       "      <th>269</th>\n",
       "      <th>270</th>\n",
       "      <th>271</th>\n",
       "      <th>272</th>\n",
       "      <th>273</th>\n",
       "      <th>274</th>\n",
       "      <th>275</th>\n",
       "      <th>276</th>\n",
       "      <th>277</th>\n",
       "      <th>278</th>\n",
       "      <th>279</th>\n",
       "      <th>280</th>\n",
       "      <th>281</th>\n",
       "      <th>282</th>\n",
       "      <th>283</th>\n",
       "      <th>284</th>\n",
       "      <th>285</th>\n",
       "      <th>286</th>\n",
       "      <th>287</th>\n",
       "      <th>288</th>\n",
       "      <th>289</th>\n",
       "      <th>290</th>\n",
       "      <th>291</th>\n",
       "      <th>292</th>\n",
       "      <th>293</th>\n",
       "      <th>294</th>\n",
       "      <th>295</th>\n",
       "      <th>296</th>\n",
       "      <th>297</th>\n",
       "      <th>298</th>\n",
       "      <th>299</th>\n",
       "      <th>300</th>\n",
       "      <th>301</th>\n",
       "      <th>302</th>\n",
       "      <th>303</th>\n",
       "      <th>304</th>\n",
       "      <th>305</th>\n",
       "      <th>306</th>\n",
       "      <th>307</th>\n",
       "      <th>308</th>\n",
       "      <th>309</th>\n",
       "      <th>310</th>\n",
       "      <th>311</th>\n",
       "      <th>312</th>\n",
       "      <th>313</th>\n",
       "      <th>314</th>\n",
       "      <th>315</th>\n",
       "      <th>316</th>\n",
       "      <th>317</th>\n",
       "      <th>318</th>\n",
       "      <th>319</th>\n",
       "      <th>320</th>\n",
       "      <th>321</th>\n",
       "      <th>322</th>\n",
       "      <th>323</th>\n",
       "      <th>324</th>\n",
       "      <th>325</th>\n",
       "      <th>326</th>\n",
       "      <th>327</th>\n",
       "      <th>328</th>\n",
       "      <th>329</th>\n",
       "      <th>330</th>\n",
       "      <th>331</th>\n",
       "      <th>332</th>\n",
       "      <th>333</th>\n",
       "      <th>334</th>\n",
       "      <th>335</th>\n",
       "      <th>336</th>\n",
       "      <th>337</th>\n",
       "      <th>338</th>\n",
       "      <th>339</th>\n",
       "      <th>340</th>\n",
       "      <th>341</th>\n",
       "      <th>342</th>\n",
       "      <th>343</th>\n",
       "      <th>344</th>\n",
       "      <th>345</th>\n",
       "      <th>346</th>\n",
       "      <th>347</th>\n",
       "      <th>348</th>\n",
       "      <th>349</th>\n",
       "      <th>350</th>\n",
       "      <th>351</th>\n",
       "      <th>352</th>\n",
       "      <th>353</th>\n",
       "      <th>354</th>\n",
       "      <th>355</th>\n",
       "      <th>356</th>\n",
       "      <th>357</th>\n",
       "      <th>358</th>\n",
       "      <th>359</th>\n",
       "      <th>360</th>\n",
       "      <th>361</th>\n",
       "      <th>362</th>\n",
       "      <th>363</th>\n",
       "      <th>364</th>\n",
       "      <th>365</th>\n",
       "      <th>366</th>\n",
       "      <th>367</th>\n",
       "      <th>368</th>\n",
       "      <th>369</th>\n",
       "      <th>370</th>\n",
       "      <th>371</th>\n",
       "      <th>372</th>\n",
       "      <th>373</th>\n",
       "      <th>374</th>\n",
       "      <th>375</th>\n",
       "      <th>376</th>\n",
       "      <th>377</th>\n",
       "      <th>378</th>\n",
       "      <th>379</th>\n",
       "      <th>380</th>\n",
       "      <th>381</th>\n",
       "      <th>382</th>\n",
       "      <th>383</th>\n",
       "      <th>384</th>\n",
       "      <th>385</th>\n",
       "      <th>386</th>\n",
       "      <th>387</th>\n",
       "      <th>388</th>\n",
       "      <th>389</th>\n",
       "      <th>390</th>\n",
       "      <th>391</th>\n",
       "      <th>392</th>\n",
       "      <th>393</th>\n",
       "      <th>394</th>\n",
       "      <th>395</th>\n",
       "      <th>396</th>\n",
       "      <th>397</th>\n",
       "      <th>398</th>\n",
       "      <th>399</th>\n",
       "      <th>400</th>\n",
       "      <th>401</th>\n",
       "      <th>402</th>\n",
       "      <th>403</th>\n",
       "      <th>404</th>\n",
       "      <th>405</th>\n",
       "      <th>406</th>\n",
       "      <th>407</th>\n",
       "      <th>408</th>\n",
       "      <th>409</th>\n",
       "      <th>410</th>\n",
       "      <th>411</th>\n",
       "      <th>412</th>\n",
       "      <th>413</th>\n",
       "      <th>414</th>\n",
       "      <th>415</th>\n",
       "      <th>416</th>\n",
       "      <th>417</th>\n",
       "      <th>418</th>\n",
       "      <th>419</th>\n",
       "      <th>420</th>\n",
       "      <th>421</th>\n",
       "      <th>422</th>\n",
       "      <th>423</th>\n",
       "      <th>424</th>\n",
       "      <th>425</th>\n",
       "      <th>426</th>\n",
       "      <th>427</th>\n",
       "      <th>428</th>\n",
       "      <th>429</th>\n",
       "      <th>430</th>\n",
       "      <th>431</th>\n",
       "      <th>432</th>\n",
       "      <th>433</th>\n",
       "      <th>434</th>\n",
       "      <th>435</th>\n",
       "      <th>436</th>\n",
       "      <th>437</th>\n",
       "      <th>438</th>\n",
       "      <th>439</th>\n",
       "      <th>440</th>\n",
       "      <th>441</th>\n",
       "      <th>442</th>\n",
       "      <th>443</th>\n",
       "      <th>444</th>\n",
       "      <th>445</th>\n",
       "      <th>446</th>\n",
       "      <th>447</th>\n",
       "      <th>448</th>\n",
       "      <th>449</th>\n",
       "      <th>450</th>\n",
       "      <th>451</th>\n",
       "      <th>452</th>\n",
       "      <th>453</th>\n",
       "      <th>454</th>\n",
       "      <th>455</th>\n",
       "      <th>456</th>\n",
       "      <th>457</th>\n",
       "      <th>458</th>\n",
       "      <th>459</th>\n",
       "      <th>460</th>\n",
       "      <th>461</th>\n",
       "      <th>462</th>\n",
       "      <th>463</th>\n",
       "      <th>464</th>\n",
       "      <th>465</th>\n",
       "      <th>466</th>\n",
       "      <th>467</th>\n",
       "      <th>468</th>\n",
       "      <th>469</th>\n",
       "      <th>470</th>\n",
       "      <th>471</th>\n",
       "      <th>472</th>\n",
       "      <th>473</th>\n",
       "      <th>474</th>\n",
       "      <th>475</th>\n",
       "      <th>476</th>\n",
       "      <th>477</th>\n",
       "      <th>478</th>\n",
       "      <th>479</th>\n",
       "      <th>480</th>\n",
       "      <th>481</th>\n",
       "      <th>482</th>\n",
       "      <th>483</th>\n",
       "      <th>484</th>\n",
       "      <th>485</th>\n",
       "      <th>486</th>\n",
       "      <th>487</th>\n",
       "      <th>488</th>\n",
       "      <th>489</th>\n",
       "      <th>490</th>\n",
       "      <th>491</th>\n",
       "      <th>492</th>\n",
       "      <th>493</th>\n",
       "      <th>494</th>\n",
       "      <th>495</th>\n",
       "      <th>496</th>\n",
       "      <th>497</th>\n",
       "      <th>498</th>\n",
       "      <th>499</th>\n",
       "      <th>500</th>\n",
       "      <th>501</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",
       "      <th>512</th>\n",
       "      <th>513</th>\n",
       "      <th>514</th>\n",
       "      <th>515</th>\n",
       "      <th>516</th>\n",
       "      <th>517</th>\n",
       "      <th>518</th>\n",
       "      <th>519</th>\n",
       "      <th>520</th>\n",
       "      <th>521</th>\n",
       "      <th>522</th>\n",
       "      <th>523</th>\n",
       "      <th>524</th>\n",
       "      <th>525</th>\n",
       "      <th>526</th>\n",
       "      <th>527</th>\n",
       "      <th>528</th>\n",
       "      <th>529</th>\n",
       "      <th>530</th>\n",
       "      <th>531</th>\n",
       "      <th>532</th>\n",
       "      <th>533</th>\n",
       "      <th>534</th>\n",
       "      <th>535</th>\n",
       "      <th>536</th>\n",
       "      <th>537</th>\n",
       "      <th>538</th>\n",
       "      <th>539</th>\n",
       "      <th>540</th>\n",
       "      <th>541</th>\n",
       "      <th>542</th>\n",
       "      <th>543</th>\n",
       "      <th>544</th>\n",
       "      <th>545</th>\n",
       "      <th>546</th>\n",
       "      <th>547</th>\n",
       "      <th>548</th>\n",
       "      <th>549</th>\n",
       "      <th>550</th>\n",
       "      <th>551</th>\n",
       "      <th>552</th>\n",
       "      <th>553</th>\n",
       "      <th>554</th>\n",
       "      <th>555</th>\n",
       "      <th>556</th>\n",
       "      <th>557</th>\n",
       "      <th>558</th>\n",
       "      <th>559</th>\n",
       "      <th>560</th>\n",
       "      <th>561</th>\n",
       "      <th>562</th>\n",
       "      <th>563</th>\n",
       "      <th>564</th>\n",
       "      <th>565</th>\n",
       "      <th>566</th>\n",
       "      <th>567</th>\n",
       "      <th>568</th>\n",
       "      <th>569</th>\n",
       "      <th>570</th>\n",
       "      <th>571</th>\n",
       "      <th>572</th>\n",
       "      <th>573</th>\n",
       "      <th>574</th>\n",
       "      <th>575</th>\n",
       "      <th>576</th>\n",
       "      <th>577</th>\n",
       "      <th>578</th>\n",
       "      <th>579</th>\n",
       "      <th>580</th>\n",
       "      <th>581</th>\n",
       "      <th>582</th>\n",
       "      <th>583</th>\n",
       "      <th>584</th>\n",
       "      <th>585</th>\n",
       "      <th>586</th>\n",
       "      <th>587</th>\n",
       "      <th>588</th>\n",
       "      <th>589</th>\n",
       "      <th>590</th>\n",
       "      <th>591</th>\n",
       "      <th>592</th>\n",
       "      <th>593</th>\n",
       "      <th>594</th>\n",
       "      <th>595</th>\n",
       "      <th>596</th>\n",
       "      <th>597</th>\n",
       "      <th>598</th>\n",
       "      <th>599</th>\n",
       "      <th>600</th>\n",
       "      <th>601</th>\n",
       "      <th>602</th>\n",
       "      <th>603</th>\n",
       "      <th>604</th>\n",
       "      <th>605</th>\n",
       "      <th>606</th>\n",
       "      <th>607</th>\n",
       "      <th>608</th>\n",
       "      <th>609</th>\n",
       "      <th>610</th>\n",
       "      <th>611</th>\n",
       "      <th>612</th>\n",
       "      <th>613</th>\n",
       "      <th>614</th>\n",
       "      <th>615</th>\n",
       "      <th>616</th>\n",
       "      <th>617</th>\n",
       "      <th>618</th>\n",
       "      <th>619</th>\n",
       "      <th>620</th>\n",
       "      <th>621</th>\n",
       "      <th>622</th>\n",
       "      <th>623</th>\n",
       "      <th>624</th>\n",
       "      <th>625</th>\n",
       "      <th>626</th>\n",
       "      <th>627</th>\n",
       "      <th>628</th>\n",
       "      <th>629</th>\n",
       "      <th>630</th>\n",
       "      <th>631</th>\n",
       "      <th>632</th>\n",
       "      <th>633</th>\n",
       "      <th>634</th>\n",
       "      <th>635</th>\n",
       "      <th>636</th>\n",
       "      <th>637</th>\n",
       "      <th>638</th>\n",
       "      <th>639</th>\n",
       "      <th>640</th>\n",
       "      <th>641</th>\n",
       "      <th>642</th>\n",
       "      <th>643</th>\n",
       "      <th>644</th>\n",
       "      <th>645</th>\n",
       "      <th>646</th>\n",
       "      <th>647</th>\n",
       "      <th>648</th>\n",
       "      <th>649</th>\n",
       "      <th>650</th>\n",
       "      <th>651</th>\n",
       "      <th>652</th>\n",
       "      <th>653</th>\n",
       "      <th>654</th>\n",
       "      <th>655</th>\n",
       "      <th>656</th>\n",
       "      <th>657</th>\n",
       "      <th>658</th>\n",
       "      <th>659</th>\n",
       "      <th>660</th>\n",
       "      <th>661</th>\n",
       "      <th>662</th>\n",
       "      <th>663</th>\n",
       "      <th>664</th>\n",
       "      <th>665</th>\n",
       "      <th>666</th>\n",
       "      <th>667</th>\n",
       "      <th>668</th>\n",
       "      <th>669</th>\n",
       "      <th>670</th>\n",
       "      <th>671</th>\n",
       "      <th>672</th>\n",
       "      <th>673</th>\n",
       "      <th>674</th>\n",
       "      <th>675</th>\n",
       "      <th>676</th>\n",
       "      <th>677</th>\n",
       "      <th>678</th>\n",
       "      <th>679</th>\n",
       "      <th>680</th>\n",
       "      <th>681</th>\n",
       "      <th>682</th>\n",
       "      <th>683</th>\n",
       "      <th>684</th>\n",
       "      <th>685</th>\n",
       "      <th>686</th>\n",
       "      <th>687</th>\n",
       "      <th>688</th>\n",
       "      <th>689</th>\n",
       "      <th>690</th>\n",
       "      <th>691</th>\n",
       "      <th>692</th>\n",
       "      <th>693</th>\n",
       "      <th>694</th>\n",
       "      <th>695</th>\n",
       "      <th>696</th>\n",
       "      <th>697</th>\n",
       "      <th>698</th>\n",
       "      <th>699</th>\n",
       "      <th>700</th>\n",
       "      <th>701</th>\n",
       "      <th>702</th>\n",
       "      <th>703</th>\n",
       "      <th>704</th>\n",
       "      <th>705</th>\n",
       "      <th>706</th>\n",
       "      <th>707</th>\n",
       "      <th>708</th>\n",
       "      <th>709</th>\n",
       "      <th>710</th>\n",
       "      <th>711</th>\n",
       "      <th>712</th>\n",
       "      <th>713</th>\n",
       "      <th>714</th>\n",
       "      <th>715</th>\n",
       "      <th>716</th>\n",
       "      <th>717</th>\n",
       "      <th>718</th>\n",
       "      <th>719</th>\n",
       "      <th>720</th>\n",
       "      <th>721</th>\n",
       "      <th>722</th>\n",
       "      <th>723</th>\n",
       "      <th>724</th>\n",
       "      <th>725</th>\n",
       "      <th>726</th>\n",
       "      <th>727</th>\n",
       "      <th>728</th>\n",
       "      <th>729</th>\n",
       "      <th>730</th>\n",
       "      <th>731</th>\n",
       "      <th>732</th>\n",
       "      <th>733</th>\n",
       "      <th>734</th>\n",
       "      <th>735</th>\n",
       "      <th>736</th>\n",
       "      <th>737</th>\n",
       "      <th>738</th>\n",
       "      <th>739</th>\n",
       "      <th>740</th>\n",
       "      <th>741</th>\n",
       "      <th>742</th>\n",
       "      <th>743</th>\n",
       "      <th>744</th>\n",
       "      <th>745</th>\n",
       "      <th>746</th>\n",
       "      <th>747</th>\n",
       "      <th>748</th>\n",
       "      <th>749</th>\n",
       "      <th>750</th>\n",
       "      <th>751</th>\n",
       "      <th>752</th>\n",
       "      <th>753</th>\n",
       "      <th>754</th>\n",
       "      <th>755</th>\n",
       "      <th>756</th>\n",
       "      <th>757</th>\n",
       "      <th>758</th>\n",
       "      <th>759</th>\n",
       "      <th>760</th>\n",
       "      <th>761</th>\n",
       "      <th>762</th>\n",
       "      <th>763</th>\n",
       "      <th>764</th>\n",
       "      <th>765</th>\n",
       "      <th>766</th>\n",
       "      <th>767</th>\n",
       "      <th>768</th>\n",
       "      <th>769</th>\n",
       "      <th>770</th>\n",
       "      <th>771</th>\n",
       "      <th>772</th>\n",
       "      <th>773</th>\n",
       "      <th>774</th>\n",
       "      <th>775</th>\n",
       "      <th>776</th>\n",
       "      <th>777</th>\n",
       "      <th>778</th>\n",
       "      <th>779</th>\n",
       "      <th>780</th>\n",
       "      <th>781</th>\n",
       "      <th>782</th>\n",
       "      <th>783</th>\n",
       "      <th>784</th>\n",
       "      <th>785</th>\n",
       "      <th>786</th>\n",
       "      <th>787</th>\n",
       "      <th>788</th>\n",
       "      <th>789</th>\n",
       "      <th>790</th>\n",
       "      <th>791</th>\n",
       "      <th>792</th>\n",
       "      <th>793</th>\n",
       "      <th>794</th>\n",
       "      <th>795</th>\n",
       "      <th>796</th>\n",
       "      <th>797</th>\n",
       "      <th>798</th>\n",
       "      <th>799</th>\n",
       "      <th>800</th>\n",
       "      <th>801</th>\n",
       "      <th>802</th>\n",
       "      <th>803</th>\n",
       "      <th>804</th>\n",
       "      <th>805</th>\n",
       "      <th>806</th>\n",
       "      <th>807</th>\n",
       "      <th>808</th>\n",
       "      <th>809</th>\n",
       "      <th>810</th>\n",
       "      <th>811</th>\n",
       "      <th>812</th>\n",
       "      <th>813</th>\n",
       "      <th>814</th>\n",
       "      <th>815</th>\n",
       "      <th>816</th>\n",
       "      <th>817</th>\n",
       "      <th>818</th>\n",
       "      <th>819</th>\n",
       "      <th>820</th>\n",
       "      <th>821</th>\n",
       "      <th>822</th>\n",
       "      <th>823</th>\n",
       "      <th>824</th>\n",
       "      <th>825</th>\n",
       "      <th>826</th>\n",
       "      <th>827</th>\n",
       "      <th>828</th>\n",
       "      <th>829</th>\n",
       "      <th>830</th>\n",
       "      <th>831</th>\n",
       "      <th>832</th>\n",
       "      <th>833</th>\n",
       "      <th>834</th>\n",
       "      <th>835</th>\n",
       "      <th>836</th>\n",
       "      <th>837</th>\n",
       "      <th>838</th>\n",
       "      <th>839</th>\n",
       "      <th>840</th>\n",
       "      <th>841</th>\n",
       "      <th>842</th>\n",
       "      <th>843</th>\n",
       "      <th>844</th>\n",
       "      <th>845</th>\n",
       "      <th>846</th>\n",
       "      <th>847</th>\n",
       "      <th>848</th>\n",
       "      <th>849</th>\n",
       "      <th>850</th>\n",
       "      <th>851</th>\n",
       "      <th>852</th>\n",
       "      <th>853</th>\n",
       "      <th>854</th>\n",
       "      <th>855</th>\n",
       "      <th>856</th>\n",
       "      <th>857</th>\n",
       "      <th>858</th>\n",
       "      <th>859</th>\n",
       "      <th>860</th>\n",
       "      <th>861</th>\n",
       "      <th>862</th>\n",
       "      <th>863</th>\n",
       "      <th>864</th>\n",
       "      <th>865</th>\n",
       "      <th>866</th>\n",
       "      <th>867</th>\n",
       "      <th>868</th>\n",
       "      <th>869</th>\n",
       "      <th>870</th>\n",
       "      <th>871</th>\n",
       "      <th>872</th>\n",
       "      <th>873</th>\n",
       "      <th>874</th>\n",
       "      <th>875</th>\n",
       "      <th>876</th>\n",
       "      <th>877</th>\n",
       "      <th>878</th>\n",
       "      <th>879</th>\n",
       "      <th>880</th>\n",
       "      <th>881</th>\n",
       "      <th>882</th>\n",
       "      <th>883</th>\n",
       "      <th>884</th>\n",
       "      <th>885</th>\n",
       "      <th>886</th>\n",
       "      <th>887</th>\n",
       "      <th>888</th>\n",
       "      <th>889</th>\n",
       "      <th>890</th>\n",
       "      <th>891</th>\n",
       "      <th>892</th>\n",
       "      <th>893</th>\n",
       "      <th>894</th>\n",
       "      <th>895</th>\n",
       "      <th>896</th>\n",
       "      <th>897</th>\n",
       "      <th>898</th>\n",
       "      <th>899</th>\n",
       "      <th>900</th>\n",
       "      <th>901</th>\n",
       "      <th>902</th>\n",
       "      <th>903</th>\n",
       "      <th>904</th>\n",
       "      <th>905</th>\n",
       "      <th>906</th>\n",
       "      <th>907</th>\n",
       "      <th>908</th>\n",
       "      <th>909</th>\n",
       "      <th>910</th>\n",
       "      <th>911</th>\n",
       "      <th>912</th>\n",
       "      <th>913</th>\n",
       "      <th>914</th>\n",
       "      <th>915</th>\n",
       "      <th>916</th>\n",
       "      <th>917</th>\n",
       "      <th>918</th>\n",
       "      <th>919</th>\n",
       "      <th>920</th>\n",
       "      <th>921</th>\n",
       "      <th>922</th>\n",
       "      <th>923</th>\n",
       "      <th>924</th>\n",
       "      <th>925</th>\n",
       "      <th>926</th>\n",
       "      <th>927</th>\n",
       "      <th>928</th>\n",
       "      <th>929</th>\n",
       "      <th>930</th>\n",
       "      <th>931</th>\n",
       "      <th>932</th>\n",
       "      <th>933</th>\n",
       "      <th>934</th>\n",
       "      <th>935</th>\n",
       "      <th>936</th>\n",
       "      <th>937</th>\n",
       "      <th>938</th>\n",
       "      <th>939</th>\n",
       "      <th>940</th>\n",
       "      <th>941</th>\n",
       "      <th>942</th>\n",
       "      <th>943</th>\n",
       "      <th>944</th>\n",
       "      <th>945</th>\n",
       "      <th>946</th>\n",
       "      <th>947</th>\n",
       "      <th>948</th>\n",
       "      <th>949</th>\n",
       "      <th>950</th>\n",
       "      <th>951</th>\n",
       "      <th>952</th>\n",
       "      <th>953</th>\n",
       "      <th>954</th>\n",
       "      <th>955</th>\n",
       "      <th>956</th>\n",
       "      <th>957</th>\n",
       "      <th>958</th>\n",
       "      <th>959</th>\n",
       "      <th>960</th>\n",
       "      <th>961</th>\n",
       "      <th>962</th>\n",
       "      <th>963</th>\n",
       "      <th>964</th>\n",
       "      <th>965</th>\n",
       "      <th>966</th>\n",
       "      <th>967</th>\n",
       "      <th>968</th>\n",
       "      <th>969</th>\n",
       "      <th>970</th>\n",
       "      <th>971</th>\n",
       "      <th>972</th>\n",
       "      <th>973</th>\n",
       "      <th>974</th>\n",
       "      <th>975</th>\n",
       "      <th>976</th>\n",
       "      <th>977</th>\n",
       "      <th>978</th>\n",
       "      <th>979</th>\n",
       "      <th>980</th>\n",
       "      <th>981</th>\n",
       "      <th>982</th>\n",
       "      <th>983</th>\n",
       "      <th>984</th>\n",
       "      <th>985</th>\n",
       "      <th>986</th>\n",
       "      <th>987</th>\n",
       "      <th>988</th>\n",
       "      <th>989</th>\n",
       "      <th>990</th>\n",
       "      <th>991</th>\n",
       "      <th>992</th>\n",
       "      <th>993</th>\n",
       "      <th>994</th>\n",
       "      <th>995</th>\n",
       "      <th>996</th>\n",
       "      <th>997</th>\n",
       "      <th>998</th>\n",
       "      <th>999</th>\n",
       "      <th>1000</th>\n",
       "      <th>1001</th>\n",
       "      <th>1002</th>\n",
       "      <th>1003</th>\n",
       "      <th>1004</th>\n",
       "      <th>1005</th>\n",
       "      <th>1006</th>\n",
       "      <th>1007</th>\n",
       "      <th>1008</th>\n",
       "      <th>1009</th>\n",
       "      <th>1010</th>\n",
       "      <th>1011</th>\n",
       "      <th>1012</th>\n",
       "      <th>1013</th>\n",
       "      <th>1014</th>\n",
       "      <th>1015</th>\n",
       "      <th>1016</th>\n",
       "      <th>1017</th>\n",
       "      <th>1018</th>\n",
       "      <th>1019</th>\n",
       "      <th>1020</th>\n",
       "      <th>1021</th>\n",
       "      <th>1022</th>\n",
       "      <th>1023</th>\n",
       "      <th>1024</th>\n",
       "      <th>1025</th>\n",
       "      <th>1026</th>\n",
       "      <th>1027</th>\n",
       "      <th>1028</th>\n",
       "      <th>1029</th>\n",
       "      <th>1030</th>\n",
       "      <th>1031</th>\n",
       "      <th>1032</th>\n",
       "      <th>1033</th>\n",
       "      <th>1034</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>able</td>\n",
       "      <td>about</td>\n",
       "      <td>above</td>\n",
       "      <td>absolutely</td>\n",
       "      <td>acoustic</td>\n",
       "      <td>across</td>\n",
       "      <td>action</td>\n",
       "      <td>actually</td>\n",
       "      <td>adapter</td>\n",
       "      <td>add</td>\n",
       "      <td>added</td>\n",
       "      <td>adds</td>\n",
       "      <td>adjustment</td>\n",
       "      <td>advertised</td>\n",
       "      <td>affordable</td>\n",
       "      <td>after</td>\n",
       "      <td>again</td>\n",
       "      <td>against</td>\n",
       "      <td>age</td>\n",
       "      <td>ago</td>\n",
       "      <td>agree</td>\n",
       "      <td>all</td>\n",
       "      <td>almost</td>\n",
       "      <td>alone</td>\n",
       "      <td>along</td>\n",
       "      <td>already</td>\n",
       "      <td>also</td>\n",
       "      <td>alternatives</td>\n",
       "      <td>although</td>\n",
       "      <td>always</td>\n",
       "      <td>am</td>\n",
       "      <td>amazing</td>\n",
       "      <td>amazon</td>\n",
       "      <td>amount</td>\n",
       "      <td>amp</td>\n",
       "      <td>amps</td>\n",
       "      <td>an</td>\n",
       "      <td>and</td>\n",
       "      <td>angle</td>\n",
       "      <td>another</td>\n",
       "      <td>answer</td>\n",
       "      <td>any</td>\n",
       "      <td>anyone</td>\n",
       "      <td>anything</td>\n",
       "      <td>anyway</td>\n",
       "      <td>apart</td>\n",
       "      <td>appear</td>\n",
       "      <td>appears</td>\n",
       "      <td>apply</td>\n",
       "      <td>are</td>\n",
       "      <td>arent</td>\n",
       "      <td>arm</td>\n",
       "      <td>around</td>\n",
       "      <td>arrived</td>\n",
       "      <td>as</td>\n",
       "      <td>ask</td>\n",
       "      <td>at</td>\n",
       "      <td>attach</td>\n",
       "      <td>attack</td>\n",
       "      <td>audio</td>\n",
       "      <td>available</td>\n",
       "      <td>aware</td>\n",
       "      <td>away</td>\n",
       "      <td>awesome</td>\n",
       "      <td>awhile</td>\n",
       "      <td>awkward</td>\n",
       "      <td>back</td>\n",
       "      <td>backup</td>\n",
       "      <td>bad</td>\n",
       "      <td>bag</td>\n",
       "      <td>ball</td>\n",
       "      <td>band</td>\n",
       "      <td>bar</td>\n",
       "      <td>base</td>\n",
       "      <td>based</td>\n",
       "      <td>basic</td>\n",
       "      <td>basically</td>\n",
       "      <td>bass</td>\n",
       "      <td>be</td>\n",
       "      <td>beat</td>\n",
       "      <td>because</td>\n",
       "      <td>become</td>\n",
       "      <td>been</td>\n",
       "      <td>before</td>\n",
       "      <td>begin</td>\n",
       "      <td>behringer</td>\n",
       "      <td>being</td>\n",
       "      <td>believe</td>\n",
       "      <td>below</td>\n",
       "      <td>bend</td>\n",
       "      <td>besides</td>\n",
       "      <td>best</td>\n",
       "      <td>better</td>\n",
       "      <td>between</td>\n",
       "      <td>big</td>\n",
       "      <td>bigger</td>\n",
       "      <td>biggest</td>\n",
       "      <td>bit</td>\n",
       "      <td>black</td>\n",
       "      <td>board</td>\n",
       "      <td>body</td>\n",
       "      <td>boost</td>\n",
       "      <td>boss</td>\n",
       "      <td>both</td>\n",
       "      <td>bother</td>\n",
       "      <td>bottom</td>\n",
       "      <td>bought</td>\n",
       "      <td>box</td>\n",
       "      <td>brand</td>\n",
       "      <td>brands</td>\n",
       "      <td>break</td>\n",
       "      <td>bridge</td>\n",
       "      <td>bright</td>\n",
       "      <td>brighter</td>\n",
       "      <td>broke</td>\n",
       "      <td>broken</td>\n",
       "      <td>bronze</td>\n",
       "      <td>bucks</td>\n",
       "      <td>budget</td>\n",
       "      <td>build</td>\n",
       "      <td>built</td>\n",
       "      <td>bulky</td>\n",
       "      <td>business</td>\n",
       "      <td>but</td>\n",
       "      <td>button</td>\n",
       "      <td>buttons</td>\n",
       "      <td>buy</td>\n",
       "      <td>buying</td>\n",
       "      <td>by</td>\n",
       "      <td>bypass</td>\n",
       "      <td>cable</td>\n",
       "      <td>cables</td>\n",
       "      <td>came</td>\n",
       "      <td>can</td>\n",
       "      <td>cannot</td>\n",
       "      <td>cant</td>\n",
       "      <td>capo</td>\n",
       "      <td>capos</td>\n",
       "      <td>care</td>\n",
       "      <td>careful</td>\n",
       "      <td>case</td>\n",
       "      <td>cause</td>\n",
       "      <td>center</td>\n",
       "      <td>certain</td>\n",
       "      <td>certainly</td>\n",
       "      <td>chance</td>\n",
       "      <td>change</td>\n",
       "      <td>changed</td>\n",
       "      <td>changes</td>\n",
       "      <td>changing</td>\n",
       "      <td>cheap</td>\n",
       "      <td>cheaper</td>\n",
       "      <td>check</td>\n",
       "      <td>choice</td>\n",
       "      <td>chords</td>\n",
       "      <td>clamp</td>\n",
       "      <td>clarity</td>\n",
       "      <td>classic</td>\n",
       "      <td>clean</td>\n",
       "      <td>cleaner</td>\n",
       "      <td>clear</td>\n",
       "      <td>click</td>\n",
       "      <td>clip</td>\n",
       "      <td>clipper</td>\n",
       "      <td>close</td>\n",
       "      <td>cloth</td>\n",
       "      <td>coated</td>\n",
       "      <td>coating</td>\n",
       "      <td>color</td>\n",
       "      <td>colors</td>\n",
       "      <td>come</td>\n",
       "      <td>comes</td>\n",
       "      <td>comfortable</td>\n",
       "      <td>coming</td>\n",
       "      <td>common</td>\n",
       "      <td>company</td>\n",
       "      <td>compared</td>\n",
       "      <td>comparison</td>\n",
       "      <td>complaint</td>\n",
       "      <td>complaints</td>\n",
       "      <td>completely</td>\n",
       "      <td>compressor</td>\n",
       "      <td>condition</td>\n",
       "      <td>connect</td>\n",
       "      <td>connection</td>\n",
       "      <td>connector</td>\n",
       "      <td>consider</td>\n",
       "      <td>considering</td>\n",
       "      <td>construction</td>\n",
       "      <td>continue</td>\n",
       "      <td>control</td>\n",
       "      <td>controls</td>\n",
       "      <td>convenient</td>\n",
       "      <td>cool</td>\n",
       "      <td>cord</td>\n",
       "      <td>cost</td>\n",
       "      <td>costs</td>\n",
       "      <td>could</td>\n",
       "      <td>couldnt</td>\n",
       "      <td>couple</td>\n",
       "      <td>course</td>\n",
       "      <td>crap</td>\n",
       "      <td>crunch</td>\n",
       "      <td>currently</td>\n",
       "      <td>custom</td>\n",
       "      <td>cut</td>\n",
       "      <td>cutter</td>\n",
       "      <td>cutting</td>\n",
       "      <td>daddario</td>\n",
       "      <td>damage</td>\n",
       "      <td>danelectro</td>\n",
       "      <td>day</td>\n",
       "      <td>days</td>\n",
       "      <td>dead</td>\n",
       "      <td>deal</td>\n",
       "      <td>decent</td>\n",
       "      <td>decide</td>\n",
       "      <td>decided</td>\n",
       "      <td>deep</td>\n",
       "      <td>definitely</td>\n",
       "      <td>degree</td>\n",
       "      <td>description</td>\n",
       "      <td>design</td>\n",
       "      <td>designed</td>\n",
       "      <td>device</td>\n",
       "      <td>did</td>\n",
       "      <td>didnt</td>\n",
       "      <td>difference</td>\n",
       "      <td>different</td>\n",
       "      <td>difficult</td>\n",
       "      <td>dirt</td>\n",
       "      <td>dirty</td>\n",
       "      <td>disappointed</td>\n",
       "      <td>distortion</td>\n",
       "      <td>do</td>\n",
       "      <td>does</td>\n",
       "      <td>doesnt</td>\n",
       "      <td>doing</td>\n",
       "      <td>dollar</td>\n",
       "      <td>dollars</td>\n",
       "      <td>done</td>\n",
       "      <td>dont</td>\n",
       "      <td>double</td>\n",
       "      <td>down</td>\n",
       "      <td>downside</td>\n",
       "      <td>drop</td>\n",
       "      <td>dry</td>\n",
       "      <td>ds</td>\n",
       "      <td>dual</td>\n",
       "      <td>due</td>\n",
       "      <td>dunlop</td>\n",
       "      <td>durability</td>\n",
       "      <td>durable</td>\n",
       "      <td>during</td>\n",
       "      <td>dynamic</td>\n",
       "      <td>each</td>\n",
       "      <td>easier</td>\n",
       "      <td>easily</td>\n",
       "      <td>easy</td>\n",
       "      <td>edit</td>\n",
       "      <td>effect</td>\n",
       "      <td>effective</td>\n",
       "      <td>effects</td>\n",
       "      <td>either</td>\n",
       "      <td>electric</td>\n",
       "      <td>elixir</td>\n",
       "      <td>elixirs</td>\n",
       "      <td>else</td>\n",
       "      <td>elsewhere</td>\n",
       "      <td>end</td>\n",
       "      <td>ended</td>\n",
       "      <td>ends</td>\n",
       "      <td>enough</td>\n",
       "      <td>entire</td>\n",
       "      <td>epiphone</td>\n",
       "      <td>eq</td>\n",
       "      <td>equipment</td>\n",
       "      <td>ernie</td>\n",
       "      <td>especially</td>\n",
       "      <td>etc</td>\n",
       "      <td>even</td>\n",
       "      <td>eventually</td>\n",
       "      <td>ever</td>\n",
       "      <td>every</td>\n",
       "      <td>everyone</td>\n",
       "      <td>everything</td>\n",
       "      <td>exactly</td>\n",
       "      <td>excellent</td>\n",
       "      <td>except</td>\n",
       "      <td>excess</td>\n",
       "      <td>expect</td>\n",
       "      <td>expected</td>\n",
       "      <td>expensive</td>\n",
       "      <td>experience</td>\n",
       "      <td>extra</td>\n",
       "      <td>extremely</td>\n",
       "      <td>eye</td>\n",
       "      <td>face</td>\n",
       "      <td>fact</td>\n",
       "      <td>fair</td>\n",
       "      <td>fairly</td>\n",
       "      <td>fall</td>\n",
       "      <td>far</td>\n",
       "      <td>fast</td>\n",
       "      <td>faster</td>\n",
       "      <td>favorite</td>\n",
       "      <td>feel</td>\n",
       "      <td>feeling</td>\n",
       "      <td>feels</td>\n",
       "      <td>feet</td>\n",
       "      <td>fell</td>\n",
       "      <td>felt</td>\n",
       "      <td>fender</td>\n",
       "      <td>few</td>\n",
       "      <td>filter</td>\n",
       "      <td>filters</td>\n",
       "      <td>finally</td>\n",
       "      <td>find</td>\n",
       "      <td>fine</td>\n",
       "      <td>finger</td>\n",
       "      <td>fingers</td>\n",
       "      <td>finish</td>\n",
       "      <td>first</td>\n",
       "      <td>fit</td>\n",
       "      <td>fits</td>\n",
       "      <td>five</td>\n",
       "      <td>fixed</td>\n",
       "      <td>flat</td>\n",
       "      <td>flimsy</td>\n",
       "      <td>floor</td>\n",
       "      <td>foam</td>\n",
       "      <td>foot</td>\n",
       "      <td>for</td>\n",
       "      <td>forever</td>\n",
       "      <td>found</td>\n",
       "      <td>free</td>\n",
       "      <td>fret</td>\n",
       "      <td>fretboard</td>\n",
       "      <td>frets</td>\n",
       "      <td>friend</td>\n",
       "      <td>from</td>\n",
       "      <td>front</td>\n",
       "      <td>full</td>\n",
       "      <td>gain</td>\n",
       "      <td>gauge</td>\n",
       "      <td>gave</td>\n",
       "      <td>gear</td>\n",
       "      <td>general</td>\n",
       "      <td>get</td>\n",
       "      <td>gets</td>\n",
       "      <td>getting</td>\n",
       "      <td>gibson</td>\n",
       "      <td>gig</td>\n",
       "      <td>gigs</td>\n",
       "      <td>give</td>\n",
       "      <td>given</td>\n",
       "      <td>gives</td>\n",
       "      <td>giving</td>\n",
       "      <td>glad</td>\n",
       "      <td>glue</td>\n",
       "      <td>go</td>\n",
       "      <td>goes</td>\n",
       "      <td>going</td>\n",
       "      <td>gold</td>\n",
       "      <td>gone</td>\n",
       "      <td>good</td>\n",
       "      <td>got</td>\n",
       "      <td>gotten</td>\n",
       "      <td>great</td>\n",
       "      <td>grip</td>\n",
       "      <td>guess</td>\n",
       "      <td>guitar</td>\n",
       "      <td>guitarist</td>\n",
       "      <td>guitarists</td>\n",
       "      <td>guitars</td>\n",
       "      <td>had</td>\n",
       "      <td>half</td>\n",
       "      <td>hand</td>\n",
       "      <td>handle</td>\n",
       "      <td>hands</td>\n",
       "      <td>happy</td>\n",
       "      <td>hard</td>\n",
       "      <td>harder</td>\n",
       "      <td>hardware</td>\n",
       "      <td>has</td>\n",
       "      <td>hate</td>\n",
       "      <td>have</td>\n",
       "      <td>havent</td>\n",
       "      <td>having</td>\n",
       "      <td>he</td>\n",
       "      <td>head</td>\n",
       "      <td>heads</td>\n",
       "      <td>hear</td>\n",
       "      <td>heard</td>\n",
       "      <td>heavier</td>\n",
       "      <td>heavy</td>\n",
       "      <td>held</td>\n",
       "      <td>help</td>\n",
       "      <td>helps</td>\n",
       "      <td>here</td>\n",
       "      <td>heres</td>\n",
       "      <td>high</td>\n",
       "      <td>higher</td>\n",
       "      <td>highly</td>\n",
       "      <td>hit</td>\n",
       "      <td>hold</td>\n",
       "      <td>holder</td>\n",
       "      <td>holding</td>\n",
       "      <td>holds</td>\n",
       "      <td>hole</td>\n",
       "      <td>holes</td>\n",
       "      <td>home</td>\n",
       "      <td>honestly</td>\n",
       "      <td>hook</td>\n",
       "      <td>hope</td>\n",
       "      <td>hosa</td>\n",
       "      <td>hours</td>\n",
       "      <td>house</td>\n",
       "      <td>how</td>\n",
       "      <td>however</td>\n",
       "      <td>humidity</td>\n",
       "      <td>id</td>\n",
       "      <td>idea</td>\n",
       "      <td>ideal</td>\n",
       "      <td>if</td>\n",
       "      <td>ill</td>\n",
       "      <td>im</td>\n",
       "      <td>imho</td>\n",
       "      <td>important</td>\n",
       "      <td>impressed</td>\n",
       "      <td>improvement</td>\n",
       "      <td>in</td>\n",
       "      <td>inch</td>\n",
       "      <td>included</td>\n",
       "      <td>including</td>\n",
       "      <td>inexpensive</td>\n",
       "      <td>input</td>\n",
       "      <td>inside</td>\n",
       "      <td>install</td>\n",
       "      <td>installed</td>\n",
       "      <td>installing</td>\n",
       "      <td>instead</td>\n",
       "      <td>instructions</td>\n",
       "      <td>instrument</td>\n",
       "      <td>instruments</td>\n",
       "      <td>intended</td>\n",
       "      <td>into</td>\n",
       "      <td>is</td>\n",
       "      <td>isnt</td>\n",
       "      <td>issue</td>\n",
       "      <td>issues</td>\n",
       "      <td>it</td>\n",
       "      <td>item</td>\n",
       "      <td>its</td>\n",
       "      <td>itself</td>\n",
       "      <td>ive</td>\n",
       "      <td>jack</td>\n",
       "      <td>jazz</td>\n",
       "      <td>job</td>\n",
       "      <td>junk</td>\n",
       "      <td>just</td>\n",
       "      <td>keep</td>\n",
       "      <td>keeping</td>\n",
       "      <td>keeps</td>\n",
       "      <td>kind</td>\n",
       "      <td>kinda</td>\n",
       "      <td>kit</td>\n",
       "      <td>knob</td>\n",
       "      <td>knobs</td>\n",
       "      <td>know</td>\n",
       "      <td>kyser</td>\n",
       "      <td>large</td>\n",
       "      <td>larger</td>\n",
       "      <td>last</td>\n",
       "      <td>later</td>\n",
       "      <td>lead</td>\n",
       "      <td>learned</td>\n",
       "      <td>least</td>\n",
       "      <td>leather</td>\n",
       "      <td>leave</td>\n",
       "      <td>left</td>\n",
       "      <td>length</td>\n",
       "      <td>les</td>\n",
       "      <td>less</td>\n",
       "      <td>let</td>\n",
       "      <td>lets</td>\n",
       "      <td>level</td>\n",
       "      <td>life</td>\n",
       "      <td>lifetime</td>\n",
       "      <td>light</td>\n",
       "      <td>lighter</td>\n",
       "      <td>lightweight</td>\n",
       "      <td>like</td>\n",
       "      <td>liked</td>\n",
       "      <td>likely</td>\n",
       "      <td>line</td>\n",
       "      <td>literally</td>\n",
       "      <td>little</td>\n",
       "      <td>live</td>\n",
       "      <td>lock</td>\n",
       "      <td>locking</td>\n",
       "      <td>locks</td>\n",
       "      <td>long</td>\n",
       "      <td>longer</td>\n",
       "      <td>look</td>\n",
       "      <td>looked</td>\n",
       "      <td>looking</td>\n",
       "      <td>looks</td>\n",
       "      <td>loop</td>\n",
       "      <td>loose</td>\n",
       "      <td>lose</td>\n",
       "      <td>losing</td>\n",
       "      <td>lot</td>\n",
       "      <td>loud</td>\n",
       "      <td>love</td>\n",
       "      <td>loved</td>\n",
       "      <td>low</td>\n",
       "      <td>lower</td>\n",
       "      <td>luck</td>\n",
       "      <td>made</td>\n",
       "      <td>main</td>\n",
       "      <td>make</td>\n",
       "      <td>makes</td>\n",
       "      <td>making</td>\n",
       "      <td>many</td>\n",
       "      <td>market</td>\n",
       "      <td>marshall</td>\n",
       "      <td>martin</td>\n",
       "      <td>martins</td>\n",
       "      <td>material</td>\n",
       "      <td>matter</td>\n",
       "      <td>may</td>\n",
       "      <td>maybe</td>\n",
       "      <td>me</td>\n",
       "      <td>mean</td>\n",
       "      <td>means</td>\n",
       "      <td>medium</td>\n",
       "      <td>mellow</td>\n",
       "      <td>mention</td>\n",
       "      <td>metal</td>\n",
       "      <td>mic</td>\n",
       "      <td>microphone</td>\n",
       "      <td>microphones</td>\n",
       "      <td>mics</td>\n",
       "      <td>middle</td>\n",
       "      <td>midi</td>\n",
       "      <td>might</td>\n",
       "      <td>mike</td>\n",
       "      <td>mind</td>\n",
       "      <td>mine</td>\n",
       "      <td>mini</td>\n",
       "      <td>minor</td>\n",
       "      <td>minutes</td>\n",
       "      <td>mixer</td>\n",
       "      <td>mm</td>\n",
       "      <td>model</td>\n",
       "      <td>models</td>\n",
       "      <td>money</td>\n",
       "      <td>monster</td>\n",
       "      <td>month</td>\n",
       "      <td>months</td>\n",
       "      <td>more</td>\n",
       "      <td>most</td>\n",
       "      <td>mostly</td>\n",
       "      <td>mount</td>\n",
       "      <td>move</td>\n",
       "      <td>moving</td>\n",
       "      <td>much</td>\n",
       "      <td>music</td>\n",
       "      <td>musicians</td>\n",
       "      <td>must</td>\n",
       "      <td>mxr</td>\n",
       "      <td>my</td>\n",
       "      <td>myself</td>\n",
       "      <td>name</td>\n",
       "      <td>near</td>\n",
       "      <td>nearly</td>\n",
       "      <td>neck</td>\n",
       "      <td>need</td>\n",
       "      <td>needed</td>\n",
       "      <td>needs</td>\n",
       "      <td>never</td>\n",
       "      <td>new</td>\n",
       "      <td>next</td>\n",
       "      <td>nice</td>\n",
       "      <td>nicely</td>\n",
       "      <td>no</td>\n",
       "      <td>noise</td>\n",
       "      <td>noisy</td>\n",
       "      <td>none</td>\n",
       "      <td>nor</td>\n",
       "      <td>normal</td>\n",
       "      <td>not</td>\n",
       "      <td>note</td>\n",
       "      <td>notes</td>\n",
       "      <td>nothing</td>\n",
       "      <td>notice</td>\n",
       "      <td>noticed</td>\n",
       "      <td>now</td>\n",
       "      <td>number</td>\n",
       "      <td>nylon</td>\n",
       "      <td>occasionally</td>\n",
       "      <td>od</td>\n",
       "      <td>of</td>\n",
       "      <td>off</td>\n",
       "      <td>often</td>\n",
       "      <td>oil</td>\n",
       "      <td>ok</td>\n",
       "      <td>okay</td>\n",
       "      <td>old</td>\n",
       "      <td>on</td>\n",
       "      <td>once</td>\n",
       "      <td>one</td>\n",
       "      <td>ones</td>\n",
       "      <td>only</td>\n",
       "      <td>onoff</td>\n",
       "      <td>onto</td>\n",
       "      <td>open</td>\n",
       "      <td>opinion</td>\n",
       "      <td>options</td>\n",
       "      <td>or</td>\n",
       "      <td>order</td>\n",
       "      <td>ordered</td>\n",
       "      <td>other</td>\n",
       "      <td>others</td>\n",
       "      <td>otherwise</td>\n",
       "      <td>our</td>\n",
       "      <td>out</td>\n",
       "      <td>output</td>\n",
       "      <td>over</td>\n",
       "      <td>overall</td>\n",
       "      <td>overdrive</td>\n",
       "      <td>overdriven</td>\n",
       "      <td>own</td>\n",
       "      <td>owned</td>\n",
       "      <td>pack</td>\n",
       "      <td>package</td>\n",
       "      <td>packs</td>\n",
       "      <td>paid</td>\n",
       "      <td>pair</td>\n",
       "      <td>part</td>\n",
       "      <td>particular</td>\n",
       "      <td>paul</td>\n",
       "      <td>pay</td>\n",
       "      <td>pedal</td>\n",
       "      <td>pedals</td>\n",
       "      <td>peg</td>\n",
       "      <td>pegs</td>\n",
       "      <td>people</td>\n",
       "      <td>perfect</td>\n",
       "      <td>perfectly</td>\n",
       "      <td>perform</td>\n",
       "      <td>performance</td>\n",
       "      <td>performing</td>\n",
       "      <td>perhaps</td>\n",
       "      <td>personal</td>\n",
       "      <td>pick</td>\n",
       "      <td>picking</td>\n",
       "      <td>picks</td>\n",
       "      <td>pickups</td>\n",
       "      <td>picture</td>\n",
       "      <td>piece</td>\n",
       "      <td>pieces</td>\n",
       "      <td>pin</td>\n",
       "      <td>pins</td>\n",
       "      <td>place</td>\n",
       "      <td>planet</td>\n",
       "      <td>plastic</td>\n",
       "      <td>play</td>\n",
       "      <td>playability</td>\n",
       "      <td>played</td>\n",
       "      <td>player</td>\n",
       "      <td>players</td>\n",
       "      <td>playing</td>\n",
       "      <td>plays</td>\n",
       "      <td>please</td>\n",
       "      <td>pleased</td>\n",
       "      <td>plenty</td>\n",
       "      <td>pliers</td>\n",
       "      <td>plug</td>\n",
       "      <td>plugged</td>\n",
       "      <td>plus</td>\n",
       "      <td>point</td>\n",
       "      <td>pop</td>\n",
       "      <td>popped</td>\n",
       "      <td>pops</td>\n",
       "      <td>popular</td>\n",
       "      <td>position</td>\n",
       "      <td>possible</td>\n",
       "      <td>power</td>\n",
       "      <td>practice</td>\n",
       "      <td>prefer</td>\n",
       "      <td>preference</td>\n",
       "      <td>press</td>\n",
       "      <td>pretty</td>\n",
       "      <td>prevent</td>\n",
       "      <td>price</td>\n",
       "      <td>priced</td>\n",
       "      <td>pro</td>\n",
       "      <td>probably</td>\n",
       "      <td>problem</td>\n",
       "      <td>problems</td>\n",
       "      <td>product</td>\n",
       "      <td>products</td>\n",
       "      <td>professional</td>\n",
       "      <td>proper</td>\n",
       "      <td>properly</td>\n",
       "      <td>pull</td>\n",
       "      <td>puller</td>\n",
       "      <td>purchase</td>\n",
       "      <td>purchased</td>\n",
       "      <td>purple</td>\n",
       "      <td>purpose</td>\n",
       "      <td>push</td>\n",
       "      <td>put</td>\n",
       "      <td>putting</td>\n",
       "      <td>quality</td>\n",
       "      <td>questions</td>\n",
       "      <td>quickly</td>\n",
       "      <td>quiet</td>\n",
       "      <td>quite</td>\n",
       "      <td>range</td>\n",
       "      <td>rarely</td>\n",
       "      <td>rather</td>\n",
       "      <td>read</td>\n",
       "      <td>reading</td>\n",
       "      <td>real</td>\n",
       "      <td>really</td>\n",
       "      <td>reason</td>\n",
       "      <td>received</td>\n",
       "      <td>recently</td>\n",
       "      <td>recommend</td>\n",
       "      <td>recommended</td>\n",
       "      <td>recording</td>\n",
       "      <td>recordings</td>\n",
       "      <td>red</td>\n",
       "      <td>regular</td>\n",
       "      <td>reliable</td>\n",
       "      <td>remove</td>\n",
       "      <td>replace</td>\n",
       "      <td>replaced</td>\n",
       "      <td>replacement</td>\n",
       "      <td>require</td>\n",
       "      <td>required</td>\n",
       "      <td>response</td>\n",
       "      <td>result</td>\n",
       "      <td>return</td>\n",
       "      <td>review</td>\n",
       "      <td>reviews</td>\n",
       "      <td>rig</td>\n",
       "      <td>right</td>\n",
       "      <td>ring</td>\n",
       "      <td>rock</td>\n",
       "      <td>room</td>\n",
       "      <td>rough</td>\n",
       "      <td>round</td>\n",
       "      <td>rubber</td>\n",
       "      <td>rugged</td>\n",
       "      <td>run</td>\n",
       "      <td>running</td>\n",
       "      <td>said</td>\n",
       "      <td>same</td>\n",
       "      <td>satisfied</td>\n",
       "      <td>save</td>\n",
       "      <td>say</td>\n",
       "      <td>says</td>\n",
       "      <td>screen</td>\n",
       "      <td>screw</td>\n",
       "      <td>screws</td>\n",
       "      <td>second</td>\n",
       "      <td>secure</td>\n",
       "      <td>see</td>\n",
       "      <td>seem</td>\n",
       "      <td>seemed</td>\n",
       "      <td>seems</td>\n",
       "      <td>seen</td>\n",
       "      <td>sell</td>\n",
       "      <td>sense</td>\n",
       "      <td>sent</td>\n",
       "      <td>separate</td>\n",
       "      <td>series</td>\n",
       "      <td>serious</td>\n",
       "      <td>set</td>\n",
       "      <td>sets</td>\n",
       "      <td>setting</td>\n",
       "      <td>settings</td>\n",
       "      <td>setup</td>\n",
       "      <td>several</td>\n",
       "      <td>shape</td>\n",
       "      <td>shipped</td>\n",
       "      <td>shipping</td>\n",
       "      <td>short</td>\n",
       "      <td>shorter</td>\n",
       "      <td>should</td>\n",
       "      <td>shouldnt</td>\n",
       "      <td>show</td>\n",
       "      <td>shows</td>\n",
       "      <td>shure</td>\n",
       "      <td>side</td>\n",
       "      <td>signal</td>\n",
       "      <td>similar</td>\n",
       "      <td>simple</td>\n",
       "      <td>simply</td>\n",
       "      <td>since</td>\n",
       "      <td>singing</td>\n",
       "      <td>single</td>\n",
       "      <td>six</td>\n",
       "      <td>size</td>\n",
       "      <td>slide</td>\n",
       "      <td>slight</td>\n",
       "      <td>slightly</td>\n",
       "      <td>slinky</td>\n",
       "      <td>slip</td>\n",
       "      <td>sm</td>\n",
       "      <td>small</td>\n",
       "      <td>smaller</td>\n",
       "      <td>smooth</td>\n",
       "      <td>sms</td>\n",
       "      <td>so</td>\n",
       "      <td>soft</td>\n",
       "      <td>sold</td>\n",
       "      <td>solid</td>\n",
       "      <td>some</td>\n",
       "      <td>someone</td>\n",
       "      <td>something</td>\n",
       "      <td>sometimes</td>\n",
       "      <td>soon</td>\n",
       "      <td>sound</td>\n",
       "      <td>sounded</td>\n",
       "      <td>sounding</td>\n",
       "      <td>sounds</td>\n",
       "      <td>special</td>\n",
       "      <td>spend</td>\n",
       "      <td>sponge</td>\n",
       "      <td>spring</td>\n",
       "      <td>stage</td>\n",
       "      <td>stand</td>\n",
       "      <td>standard</td>\n",
       "      <td>stands</td>\n",
       "      <td>star</td>\n",
       "      <td>stars</td>\n",
       "      <td>start</td>\n",
       "      <td>started</td>\n",
       "      <td>state</td>\n",
       "      <td>stated</td>\n",
       "      <td>stay</td>\n",
       "      <td>steel</td>\n",
       "      <td>step</td>\n",
       "      <td>stick</td>\n",
       "      <td>stiff</td>\n",
       "      <td>still</td>\n",
       "      <td>stock</td>\n",
       "      <td>store</td>\n",
       "      <td>straight</td>\n",
       "      <td>strap</td>\n",
       "      <td>straplocks</td>\n",
       "      <td>straps</td>\n",
       "      <td>strat</td>\n",
       "      <td>stress</td>\n",
       "      <td>string</td>\n",
       "      <td>strings</td>\n",
       "      <td>strong</td>\n",
       "      <td>strumming</td>\n",
       "      <td>studio</td>\n",
       "      <td>stuff</td>\n",
       "      <td>sturdy</td>\n",
       "      <td>style</td>\n",
       "      <td>such</td>\n",
       "      <td>super</td>\n",
       "      <td>supposed</td>\n",
       "      <td>sure</td>\n",
       "      <td>surface</td>\n",
       "      <td>surprised</td>\n",
       "      <td>suspect</td>\n",
       "      <td>sustain</td>\n",
       "      <td>sweet</td>\n",
       "      <td>switch</td>\n",
       "      <td>switched</td>\n",
       "      <td>system</td>\n",
       "      <td>take</td>\n",
       "      <td>takes</td>\n",
       "      <td>technique</td>\n",
       "      <td>tell</td>\n",
       "      <td>tend</td>\n",
       "      <td>tension</td>\n",
       "      <td>testing</td>\n",
       "      <td>th</td>\n",
       "      <td>than</td>\n",
       "      <td>that</td>\n",
       "      <td>thats</td>\n",
       "      <td>the</td>\n",
       "      <td>their</td>\n",
       "      <td>them</td>\n",
       "      <td>then</td>\n",
       "      <td>there</td>\n",
       "      <td>theres</td>\n",
       "      <td>these</td>\n",
       "      <td>they</td>\n",
       "      <td>theyre</td>\n",
       "      <td>thick</td>\n",
       "      <td>thicker</td>\n",
       "      <td>thickness</td>\n",
       "      <td>thin</td>\n",
       "      <td>thing</td>\n",
       "      <td>things</td>\n",
       "      <td>think</td>\n",
       "      <td>thinking</td>\n",
       "      <td>thinner</td>\n",
       "      <td>this</td>\n",
       "      <td>those</td>\n",
       "      <td>though</td>\n",
       "      <td>thought</td>\n",
       "      <td>three</td>\n",
       "      <td>through</td>\n",
       "      <td>tight</td>\n",
       "      <td>tighten</td>\n",
       "      <td>tighter</td>\n",
       "      <td>time</td>\n",
       "      <td>times</td>\n",
       "      <td>tiny</td>\n",
       "      <td>to</td>\n",
       "      <td>today</td>\n",
       "      <td>together</td>\n",
       "      <td>tone</td>\n",
       "      <td>tones</td>\n",
       "      <td>too</td>\n",
       "      <td>took</td>\n",
       "      <td>tool</td>\n",
       "      <td>top</td>\n",
       "      <td>tortex</td>\n",
       "      <td>totally</td>\n",
       "      <td>traditional</td>\n",
       "      <td>tried</td>\n",
       "      <td>true</td>\n",
       "      <td>try</td>\n",
       "      <td>trying</td>\n",
       "      <td>tube</td>\n",
       "      <td>tune</td>\n",
       "      <td>tuned</td>\n",
       "      <td>tuners</td>\n",
       "      <td>tuning</td>\n",
       "      <td>turn</td>\n",
       "      <td>turning</td>\n",
       "      <td>two</td>\n",
       "      <td>type</td>\n",
       "      <td>types</td>\n",
       "      <td>typically</td>\n",
       "      <td>ukulele</td>\n",
       "      <td>under</td>\n",
       "      <td>unfortunately</td>\n",
       "      <td>unique</td>\n",
       "      <td>unit</td>\n",
       "      <td>unless</td>\n",
       "      <td>until</td>\n",
       "      <td>up</td>\n",
       "      <td>update</td>\n",
       "      <td>upon</td>\n",
       "      <td>us</td>\n",
       "      <td>use</td>\n",
       "      <td>used</td>\n",
       "      <td>useful</td>\n",
       "      <td>useless</td>\n",
       "      <td>uses</td>\n",
       "      <td>using</td>\n",
       "      <td>usually</td>\n",
       "      <td>value</td>\n",
       "      <td>variety</td>\n",
       "      <td>version</td>\n",
       "      <td>very</td>\n",
       "      <td>vocal</td>\n",
       "      <td>vocals</td>\n",
       "      <td>voice</td>\n",
       "      <td>volume</td>\n",
       "      <td>vox</td>\n",
       "      <td>wah</td>\n",
       "      <td>want</td>\n",
       "      <td>wanted</td>\n",
       "      <td>warm</td>\n",
       "      <td>was</td>\n",
       "      <td>wasnt</td>\n",
       "      <td>waves</td>\n",
       "      <td>way</td>\n",
       "      <td>we</td>\n",
       "      <td>wear</td>\n",
       "      <td>week</td>\n",
       "      <td>weeks</td>\n",
       "      <td>weight</td>\n",
       "      <td>well</td>\n",
       "      <td>went</td>\n",
       "      <td>were</td>\n",
       "      <td>what</td>\n",
       "      <td>whatever</td>\n",
       "      <td>whats</td>\n",
       "      <td>when</td>\n",
       "      <td>where</td>\n",
       "      <td>which</td>\n",
       "      <td>while</td>\n",
       "      <td>white</td>\n",
       "      <td>who</td>\n",
       "      <td>whole</td>\n",
       "      <td>why</td>\n",
       "      <td>will</td>\n",
       "      <td>winder</td>\n",
       "      <td>winding</td>\n",
       "      <td>wire</td>\n",
       "      <td>wish</td>\n",
       "      <td>with</td>\n",
       "      <td>within</td>\n",
       "      <td>without</td>\n",
       "      <td>wonder</td>\n",
       "      <td>wont</td>\n",
       "      <td>wood</td>\n",
       "      <td>work</td>\n",
       "      <td>worked</td>\n",
       "      <td>working</td>\n",
       "      <td>works</td>\n",
       "      <td>world</td>\n",
       "      <td>worry</td>\n",
       "      <td>worse</td>\n",
       "      <td>worth</td>\n",
       "      <td>would</td>\n",
       "      <td>wouldnt</td>\n",
       "      <td>wound</td>\n",
       "      <td>writing</td>\n",
       "      <td>wrong</td>\n",
       "      <td>xlr</td>\n",
       "      <td>year</td>\n",
       "      <td>years</td>\n",
       "      <td>yes</td>\n",
       "      <td>yet</td>\n",
       "      <td>you</td>\n",
       "      <td>youd</td>\n",
       "      <td>youll</td>\n",
       "      <td>your</td>\n",
       "      <td>youre</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   0      1      2           3         4       5       6         7     \\\n",
       "0  able  about  above  absolutely  acoustic  across  action  actually   \n",
       "\n",
       "      8    9      10    11          12          13          14     15    \\\n",
       "0  adapter  add  added  adds  adjustment  advertised  affordable  after   \n",
       "\n",
       "    16       17   18   19     20   21      22     23     24       25    26    \\\n",
       "0  again  against  age  ago  agree  all  almost  alone  along  already  also   \n",
       "\n",
       "           27        28      29   30       31      32      33   34    35    \\\n",
       "0  alternatives  although  always   am  amazing  amazon  amount  amp  amps   \n",
       "\n",
       "  36   37     38       39      40   41      42        43      44     45    \\\n",
       "0   an  and  angle  another  answer  any  anyone  anything  anyway  apart   \n",
       "\n",
       "     46       47     48   49     50   51      52       53   54   55   56    \\\n",
       "0  appear  appears  apply  are  arent  arm  around  arrived   as  ask   at   \n",
       "\n",
       "     57      58     59         60     61    62       63      64       65    \\\n",
       "0  attach  attack  audio  available  aware  away  awesome  awhile  awkward   \n",
       "\n",
       "   66      67   68   69    70    71   72    73     74     75         76    \\\n",
       "0  back  backup  bad  bag  ball  band  bar  base  based  basic  basically   \n",
       "\n",
       "   77   78    79       80      81    82      83     84         85     86    \\\n",
       "0  bass   be  beat  because  become  been  before  begin  behringer  being   \n",
       "\n",
       "      87     88    89       90    91      92       93   94      95       96    \\\n",
       "0  believe  below  bend  besides  best  better  between  big  bigger  biggest   \n",
       "\n",
       "  97     98     99    100    101   102   103     104     105     106  107   \\\n",
       "0  bit  black  board  body  boost  boss  both  bother  bottom  bought  box   \n",
       "\n",
       "    108     109    110     111     112       113    114     115     116   \\\n",
       "0  brand  brands  break  bridge  bright  brighter  broke  broken  bronze   \n",
       "\n",
       "    117     118    119    120    121       122  123     124      125  126   \\\n",
       "0  bucks  budget  build  built  bulky  business  but  button  buttons  buy   \n",
       "\n",
       "     127  128     129    130     131   132  133     134   135   136    137   \\\n",
       "0  buying   by  bypass  cable  cables  came  can  cannot  cant  capo  capos   \n",
       "\n",
       "   138      139   140    141     142      143        144     145     146   \\\n",
       "0  care  careful  case  cause  center  certain  certainly  chance  change   \n",
       "\n",
       "      147      148       149    150      151    152     153     154    155   \\\n",
       "0  changed  changes  changing  cheap  cheaper  check  choice  chords  clamp   \n",
       "\n",
       "      156      157    158      159    160    161   162      163    164   \\\n",
       "0  clarity  classic  clean  cleaner  clear  click  clip  clipper  close   \n",
       "\n",
       "    165     166      167    168     169   170    171          172     173   \\\n",
       "0  cloth  coated  coating  color  colors  come  comes  comfortable  coming   \n",
       "\n",
       "     174      175       176         177        178         179         180   \\\n",
       "0  common  company  compared  comparison  complaint  complaints  completely   \n",
       "\n",
       "         181        182      183         184        185       186   \\\n",
       "0  compressor  condition  connect  connection  connector  consider   \n",
       "\n",
       "          187           188       189      190       191         192   193   \\\n",
       "0  considering  construction  continue  control  controls  convenient  cool   \n",
       "\n",
       "   194   195    196    197      198     199     200   201     202        203   \\\n",
       "0  cord  cost  costs  could  couldnt  couple  course  crap  crunch  currently   \n",
       "\n",
       "     204  205     206      207       208     209         210  211   212   \\\n",
       "0  custom  cut  cutter  cutting  daddario  damage  danelectro  day  days   \n",
       "\n",
       "   213   214     215     216      217   218         219     220          221   \\\n",
       "0  dead  deal  decent  decide  decided  deep  definitely  degree  description   \n",
       "\n",
       "     222       223     224  225    226         227        228        229   \\\n",
       "0  design  designed  device  did  didnt  difference  different  difficult   \n",
       "\n",
       "   230    231           232         233  234   235     236    237     238   \\\n",
       "0  dirt  dirty  disappointed  distortion   do  does  doesnt  doing  dollar   \n",
       "\n",
       "      239   240   241     242   243       244   245  246  247   248  249   \\\n",
       "0  dollars  done  dont  double  down  downside  drop  dry   ds  dual  due   \n",
       "\n",
       "     250         251      252     253      254   255     256     257   258   \\\n",
       "0  dunlop  durability  durable  during  dynamic  each  easier  easily  easy   \n",
       "\n",
       "   259     260        261      262     263       264     265      266   267   \\\n",
       "0  edit  effect  effective  effects  either  electric  elixir  elixirs  else   \n",
       "\n",
       "        268  269    270   271     272     273       274  275        276   \\\n",
       "0  elsewhere  end  ended  ends  enough  entire  epiphone   eq  equipment   \n",
       "\n",
       "    277         278  279   280         281   282    283       284   \\\n",
       "0  ernie  especially  etc  even  eventually  ever  every  everyone   \n",
       "\n",
       "         285      286        287     288     289     290       291   \\\n",
       "0  everything  exactly  excellent  except  excess  expect  expected   \n",
       "\n",
       "        292         293    294        295  296   297   298   299     300   \\\n",
       "0  expensive  experience  extra  extremely  eye  face  fact  fair  fairly   \n",
       "\n",
       "   301  302   303     304       305   306      307    308   309   310   311   \\\n",
       "0  fall  far  fast  faster  favorite  feel  feeling  feels  feet  fell  felt   \n",
       "\n",
       "     312  313     314      315      316   317   318     319      320     321   \\\n",
       "0  fender  few  filter  filters  finally  find  fine  finger  fingers  finish   \n",
       "\n",
       "    322  323   324   325    326   327     328    329   330   331  332   \\\n",
       "0  first  fit  fits  five  fixed  flat  flimsy  floor  foam  foot  for   \n",
       "\n",
       "      333    334   335   336        337    338     339   340    341   342   \\\n",
       "0  forever  found  free  fret  fretboard  frets  friend  from  front  full   \n",
       "\n",
       "   343    344   345   346      347  348   349      350     351  352   353   \\\n",
       "0  gain  gauge  gave  gear  general  get  gets  getting  gibson  gig  gigs   \n",
       "\n",
       "   354    355    356     357   358   359  360   361    362   363   364   365   \\\n",
       "0  give  given  gives  giving  glad  glue   go  goes  going  gold  gone  good   \n",
       "\n",
       "  366     367    368   369    370     371        372         373      374   \\\n",
       "0  got  gotten  great  grip  guess  guitar  guitarist  guitarists  guitars   \n",
       "\n",
       "  375   376   377     378    379    380   381     382       383  384   385   \\\n",
       "0  had  half  hand  handle  hands  happy  hard  harder  hardware  has  hate   \n",
       "\n",
       "   386     387     388  389   390    391   392    393      394    395   396   \\\n",
       "0  have  havent  having   he  head  heads  hear  heard  heavier  heavy  held   \n",
       "\n",
       "   397    398   399    400   401     402     403  404   405     406      407   \\\n",
       "0  help  helps  here  heres  high  higher  highly  hit  hold  holder  holding   \n",
       "\n",
       "    408   409    410   411       412   413   414   415    416    417  418   \\\n",
       "0  holds  hole  holes  home  honestly  hook  hope  hosa  hours  house  how   \n",
       "\n",
       "      419       420  421   422    423  424  425  426   427        428   \\\n",
       "0  however  humidity   id  idea  ideal   if  ill   im  imho  important   \n",
       "\n",
       "        429          430  431   432       433        434          435    436   \\\n",
       "0  impressed  improvement   in  inch  included  including  inexpensive  input   \n",
       "\n",
       "     437      438        439         440      441           442         443   \\\n",
       "0  inside  install  installed  installing  instead  instructions  instrument   \n",
       "\n",
       "          444       445   446  447   448    449     450  451   452  453   \\\n",
       "0  instruments  intended  into   is  isnt  issue  issues   it  item  its   \n",
       "\n",
       "     454  455   456   457  458   459   460   461      462    463   464   \\\n",
       "0  itself  ive  jack  jazz  job  junk  just  keep  keeping  keeps  kind   \n",
       "\n",
       "    465  466   467    468   469    470    471     472   473    474   475   \\\n",
       "0  kinda  kit  knob  knobs  know  kyser  large  larger  last  later  lead   \n",
       "\n",
       "      476    477      478    479   480     481  482   483  484   485    486   \\\n",
       "0  learned  least  leather  leave  left  length  les  less  let  lets  level   \n",
       "\n",
       "   487       488    489      490          491   492    493     494   495   \\\n",
       "0  life  lifetime  light  lighter  lightweight  like  liked  likely  line   \n",
       "\n",
       "        496     497   498   499      500    501   502     503   504     505   \\\n",
       "0  literally  little  live  lock  locking  locks  long  longer  look  looked   \n",
       "\n",
       "      506    507   508    509   510     511  512   513   514    515  516   \\\n",
       "0  looking  looks  loop  loose  lose  losing  lot  loud  love  loved  low   \n",
       "\n",
       "    517   518   519   520   521    522     523   524     525       526   \\\n",
       "0  lower  luck  made  main  make  makes  making  many  market  marshall   \n",
       "\n",
       "     527      528       529     530  531    532  533   534    535     536   \\\n",
       "0  martin  martins  material  matter  may  maybe   me  mean  means  medium   \n",
       "\n",
       "     537      538    539  540         541          542   543     544   545   \\\n",
       "0  mellow  mention  metal  mic  microphone  microphones  mics  middle  midi   \n",
       "\n",
       "    546   547   548   549   550    551      552    553  554    555     556   \\\n",
       "0  might  mike  mind  mine  mini  minor  minutes  mixer   mm  model  models   \n",
       "\n",
       "    557      558    559     560   561   562     563    564   565     566   \\\n",
       "0  money  monster  month  months  more  most  mostly  mount  move  moving   \n",
       "\n",
       "   567    568        569   570  571  572     573   574   575     576   577   \\\n",
       "0  much  music  musicians  must  mxr   my  myself  name  near  nearly  neck   \n",
       "\n",
       "   578     579    580    581  582   583   584     585  586    587    588   \\\n",
       "0  need  needed  needs  never  new  next  nice  nicely   no  noise  noisy   \n",
       "\n",
       "   589  590     591  592   593    594      595     596      597  598     599   \\\n",
       "0  none  nor  normal  not  note  notes  nothing  notice  noticed  now  number   \n",
       "\n",
       "    600           601  602  603  604    605  606  607   608  609  610   611   \\\n",
       "0  nylon  occasionally   od   of  off  often  oil   ok  okay  old   on  once   \n",
       "\n",
       "  612   613   614    615   616   617      618      619  620    621      622   \\\n",
       "0  one  ones  only  onoff  onto  open  opinion  options   or  order  ordered   \n",
       "\n",
       "    623     624        625  626  627     628   629      630        631   \\\n",
       "0  other  others  otherwise  our  out  output  over  overall  overdrive   \n",
       "\n",
       "         632  633    634   635      636    637   638   639   640         641   \\\n",
       "0  overdriven  own  owned  pack  package  packs  paid  pair  part  particular   \n",
       "\n",
       "   642  643    644     645  646   647     648      649        650      651   \\\n",
       "0  paul  pay  pedal  pedals  peg  pegs  people  perfect  perfectly  perform   \n",
       "\n",
       "          652         653      654       655   656      657    658      659   \\\n",
       "0  performance  performing  perhaps  personal  pick  picking  picks  pickups   \n",
       "\n",
       "      660    661     662  663   664    665     666      667   668   \\\n",
       "0  picture  piece  pieces  pin  pins  place  planet  plastic  play   \n",
       "\n",
       "          669     670     671      672      673    674     675      676   \\\n",
       "0  playability  played  player  players  playing  plays  please  pleased   \n",
       "\n",
       "     677     678   679      680   681    682  683     684   685      686   \\\n",
       "0  plenty  pliers  plug  plugged  plus  point  pop  popped  pops  popular   \n",
       "\n",
       "       687       688    689       690     691         692    693     694   \\\n",
       "0  position  possible  power  practice  prefer  preference  press  pretty   \n",
       "\n",
       "      695    696     697  698       699      700       701      702   \\\n",
       "0  prevent  price  priced  pro  probably  problem  problems  product   \n",
       "\n",
       "       703           704     705       706   707     708       709   \\\n",
       "0  products  professional  proper  properly  pull  puller  purchase   \n",
       "\n",
       "        710     711      712   713  714      715      716        717   \\\n",
       "0  purchased  purple  purpose  push  put  putting  quality  questions   \n",
       "\n",
       "      718    719    720    721     722     723   724      725   726     727   \\\n",
       "0  quickly  quiet  quite  range  rarely  rather  read  reading  real  really   \n",
       "\n",
       "     728       729       730        731          732        733         734   \\\n",
       "0  reason  received  recently  recommend  recommended  recording  recordings   \n",
       "\n",
       "  735      736       737     738      739       740          741      742   \\\n",
       "0  red  regular  reliable  remove  replace  replaced  replacement  require   \n",
       "\n",
       "       743       744     745     746     747      748  749    750   751   \\\n",
       "0  required  response  result  return  review  reviews  rig  right  ring   \n",
       "\n",
       "   752   753    754    755     756     757  758      759   760   761   \\\n",
       "0  rock  room  rough  round  rubber  rugged  run  running  said  same   \n",
       "\n",
       "        762   763  764   765     766    767     768     769     770  771   \\\n",
       "0  satisfied  save  say  says  screen  screw  screws  second  secure  see   \n",
       "\n",
       "   772     773    774   775   776    777   778       779     780      781   \\\n",
       "0  seem  seemed  seems  seen  sell  sense  sent  separate  series  serious   \n",
       "\n",
       "  782   783      784       785    786      787    788      789       790   \\\n",
       "0  set  sets  setting  settings  setup  several  shape  shipped  shipping   \n",
       "\n",
       "    791      792     793       794   795    796    797   798     799   \\\n",
       "0  short  shorter  should  shouldnt  show  shows  shure  side  signal   \n",
       "\n",
       "      800     801     802    803      804     805  806   807    808     809   \\\n",
       "0  similar  simple  simply  since  singing  single  six  size  slide  slight   \n",
       "\n",
       "       810     811   812  813    814      815     816  817  818   819   820   \\\n",
       "0  slightly  slinky  slip   sm  small  smaller  smooth  sms   so  soft  sold   \n",
       "\n",
       "    821   822      823        824        825   826    827      828       829   \\\n",
       "0  solid  some  someone  something  sometimes  soon  sound  sounded  sounding   \n",
       "\n",
       "     830      831    832     833     834    835    836       837     838   \\\n",
       "0  sounds  special  spend  sponge  spring  stage  stand  standard  stands   \n",
       "\n",
       "   839    840    841      842    843     844   845    846   847    848   \\\n",
       "0  star  stars  start  started  state  stated  stay  steel  step  stick   \n",
       "\n",
       "    849    850    851    852       853    854         855     856    857   \\\n",
       "0  stiff  still  stock  store  straight  strap  straplocks  straps  strat   \n",
       "\n",
       "     858     859      860     861        862     863    864     865    866   \\\n",
       "0  stress  string  strings  strong  strumming  studio  stuff  sturdy  style   \n",
       "\n",
       "   867    868       869   870      871        872      873      874    875   \\\n",
       "0  such  super  supposed  sure  surface  surprised  suspect  sustain  sweet   \n",
       "\n",
       "     876       877     878   879    880        881   882   883      884   \\\n",
       "0  switch  switched  system  take  takes  technique  tell  tend  tension   \n",
       "\n",
       "      885  886   887   888    889  890    891   892   893    894     895   \\\n",
       "0  testing   th  than  that  thats  the  their  them  then  there  theres   \n",
       "\n",
       "    896   897     898    899      900        901   902    903     904    905   \\\n",
       "0  these  they  theyre  thick  thicker  thickness  thin  thing  things  think   \n",
       "\n",
       "       906      907   908    909     910      911    912      913    914   \\\n",
       "0  thinking  thinner  this  those  though  thought  three  through  tight   \n",
       "\n",
       "      915      916   917    918   919  920    921       922   923    924   \\\n",
       "0  tighten  tighter  time  times  tiny   to  today  together  tone  tones   \n",
       "\n",
       "  925   926   927  928     929      930          931    932   933  934   \\\n",
       "0  too  took  tool  top  tortex  totally  traditional  tried  true  try   \n",
       "\n",
       "     935   936   937    938     939     940   941      942  943   944    945   \\\n",
       "0  trying  tube  tune  tuned  tuners  tuning  turn  turning  two  type  types   \n",
       "\n",
       "        946      947    948            949     950   951     952    953  954   \\\n",
       "0  typically  ukulele  under  unfortunately  unique  unit  unless  until   up   \n",
       "\n",
       "     955   956  957  958   959     960      961   962    963      964    965   \\\n",
       "0  update  upon   us  use  used  useful  useless  uses  using  usually  value   \n",
       "\n",
       "      966      967   968    969     970    971     972  973  974   975   \\\n",
       "0  variety  version  very  vocal  vocals  voice  volume  vox  wah  want   \n",
       "\n",
       "     976   977  978    979    980  981  982   983   984    985     986   987   \\\n",
       "0  wanted  warm  was  wasnt  waves  way   we  wear  week  weeks  weight  well   \n",
       "\n",
       "   988   989   990       991    992   993    994    995    996    997  998   \\\n",
       "0  went  were  what  whatever  whats  when  where  which  while  white  who   \n",
       "\n",
       "    999  1000  1001    1002     1003  1004  1005  1006    1007     1008  \\\n",
       "0  whole  why  will  winder  winding  wire  wish  with  within  without   \n",
       "\n",
       "     1009  1010  1011  1012    1013     1014   1015   1016   1017   1018  \\\n",
       "0  wonder  wont  wood  work  worked  working  works  world  worry  worse   \n",
       "\n",
       "    1019   1020     1021   1022     1023   1024 1025  1026   1027 1028 1029  \\\n",
       "0  worth  would  wouldnt  wound  writing  wrong  xlr  year  years  yes  yet   \n",
       "\n",
       "  1030  1031   1032  1033   1034  \n",
       "0  you  youd  youll  your  youre  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Uncertain in Test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "t = clf.predict_proba(X_test_vec)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uncertain Objects count =  33\n"
     ]
    }
   ],
   "source": [
    "count = 0\n",
    "for i in range(X_test_vec.shape[0]):\n",
    "    if t[i][0] < 0.6 and t[i][0] > 0.3 :\n",
    "            #print(t[i][0])\n",
    "            count += 1\n",
    "print(\"Uncertain Objects count = \" ,count )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 3: Active Learning "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Artificial Labelling - Chi Square to get all postive and negative words from Train dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "pos_chi_scores = np.argsort(chi2(X_train_vec,y_train)[1])[::-1]\n",
    "neg_chi_scores = np.argsort(chi2(X_train_vec,y_train)[0])[::-1]\n",
    "pos_words = []\n",
    "for i in range(features.shape[1]):\n",
    "    pos_words.append(features[pos_chi_scores[i]])\n",
    "\n",
    "neg_words = []\n",
    "for i in range(features.shape[1]):\n",
    "    neg_words.append(features[neg_chi_scores[i]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>amazon</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>683</th>\n",
       "      <td>pop</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>365</th>\n",
       "      <td>good</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>694</th>\n",
       "      <td>pretty</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>748</th>\n",
       "      <td>reviews</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>278</th>\n",
       "      <td>especially</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>690</th>\n",
       "      <td>practice</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>206</th>\n",
       "      <td>cutter</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1006</th>\n",
       "      <td>with</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>934</th>\n",
       "      <td>try</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               0\n",
       "32        amazon\n",
       "683          pop\n",
       "365         good\n",
       "694       pretty\n",
       "748      reviews\n",
       "278   especially\n",
       "690     practice\n",
       "206       cutter\n",
       "1006        with\n",
       "934          try"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(pos_words[:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>961</th>\n",
       "      <td>useless</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>545</th>\n",
       "      <td>midi</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>190</th>\n",
       "      <td>control</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>540</th>\n",
       "      <td>mic</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>558</th>\n",
       "      <td>monster</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>543</th>\n",
       "      <td>mics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>588</th>\n",
       "      <td>noisy</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>984</th>\n",
       "      <td>week</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>618</th>\n",
       "      <td>opinion</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>724</th>\n",
       "      <td>read</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           0\n",
       "961  useless\n",
       "545     midi\n",
       "190  control\n",
       "540      mic\n",
       "558  monster\n",
       "543     mics\n",
       "588    noisy\n",
       "984     week\n",
       "618  opinion\n",
       "724     read"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(neg_words[:10])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Active Learning using  Tandem Learning \n",
    "In which we mark top words from chi-square to rationale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "r = 1\n",
    "o = 0.01"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_train_rationale_vec = X_train_vec.copy() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_train_rationale_vec = X_train_rationale_vec.todense()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_train_rationale_vec = X_train_rationale_vec * o"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "for doc_i in range(X_train_rationale_vec.shape[0]):\n",
    "    if y_train[doc_i] == 1: # pos document\n",
    "        for j in range(50):\n",
    "            if X_train_rationale_vec[doc_i,pos_words[j].name] == 0.01:\n",
    "                X_train_rationale_vec[doc_i,pos_words[j].name] = 1\n",
    "                #print(pos_words[j][0])\n",
    "    if y_train[doc_i] == 0: # pos document\n",
    "        for k in range(50):\n",
    "            if X_train_rationale_vec[doc_i,pos_words[k].name] == 0.01:\n",
    "                X_train_rationale_vec[doc_i,neg_words[k].name] = 1\n",
    "        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Example for document 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Not much to write about here but it does exactly what its supposed to filters out the pop sounds now my recordings are much more crisp it is one of the lowest prices pop filters on amazon so might as well buy it they honestly work the same despite their pricing'"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.loc[0,'cleaned_review_text']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Test "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5 Words rationale  document 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.sum(X_train_rationale_vec[0] == 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 37 words as non-rationale  document 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "37"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.sum(X_train_rationale_vec[0] == 0.01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.95999999999999996"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = LogisticRegression(penalty='l2', C = 0.001)\n",
    "clf.fit(X_train_rationale_vec[:50], y_train[:50])\n",
    "y_predict = clf.predict(X_train_rationale_vec[:50])\n",
    "y_proba = clf.predict_proba(X_test_vec[:50])\n",
    "\n",
    "accuracy_score(y_test[:50], y_predict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stats  = []\n",
    "f_we_wr = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def AL_tandem_learning(size):\n",
    "    clf = LogisticRegression(penalty='l2', C = 0.001)\n",
    "    clf.fit(X_train_rationale_vec[:size], y_train[:size])\n",
    "    y_predict = clf.predict(X_test_vec[:size])\n",
    "    y_proba = clf.predict_proba(X_test_vec[:size])\n",
    "\n",
    "    a = accuracy_score(y_test[:size], y_predict)\n",
    "    \n",
    "    f_we_wr.append(top_weighted_features(clf.coef_[0], \"LR\", features))\n",
    "    preds = y_proba[:,1]\n",
    "    fpr, tpr, threshold = metrics.roc_curve(y_test[:size], preds)\n",
    "    roc_auc = metrics.auc(fpr, tpr)\n",
    "    stats.append([a,roc_auc])\n",
    "    print(f\"{size}, accuracy = {a}, auc = {roc_auc}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40, accuracy = 0.8, auc = 0.65625\n",
      "45, accuracy = 0.7111111111111111, auc = 0.5277777777777778\n",
      "50, accuracy = 0.62, auc = 0.5367647058823529\n",
      "55, accuracy = 0.45454545454545453, auc = 0.45285714285714285\n",
      "60, accuracy = 0.48333333333333334, auc = 0.4557416267942584\n",
      "65, accuracy = 0.4, auc = 0.44999999999999996\n",
      "70, accuracy = 0.4, auc = 0.45500420521446594\n",
      "75, accuracy = 0.41333333333333333, auc = 0.4662840746054519\n",
      "80, accuracy = 0.425, auc = 0.4717460317460317\n",
      "85, accuracy = 0.4235294117647059, auc = 0.4905239687848383\n",
      "90, accuracy = 0.45555555555555555, auc = 0.5007422068283027\n",
      "95, accuracy = 0.47368421052631576, auc = 0.5146406388642413\n",
      "100, accuracy = 0.49, auc = 0.5038015206082433\n",
      "105, accuracy = 0.4857142857142857, auc = 0.5108932461873639\n",
      "110, accuracy = 0.4818181818181818, auc = 0.5112433862433863\n",
      "115, accuracy = 0.5217391304347826, auc = 0.5191256830601093\n",
      "120, accuracy = 0.5333333333333333, auc = 0.514965034965035\n",
      "125, accuracy = 0.552, auc = 0.507012987012987\n",
      "130, accuracy = 0.5615384615384615, auc = 0.5108590733590734\n",
      "135, accuracy = 0.562962962962963, auc = 0.5040304523063144\n",
      "140, accuracy = 0.5642857142857143, auc = 0.5197916666666667\n",
      "145, accuracy = 0.5586206896551724, auc = 0.5019357336430508\n",
      "150, accuracy = 0.5533333333333333, auc = 0.512579185520362\n",
      "155, accuracy = 0.567741935483871, auc = 0.5074906367041199\n",
      "160, accuracy = 0.575, auc = 0.5158080564917348\n"
     ]
    }
   ],
   "source": [
    "x_ax = []\n",
    "for i in range(25):\n",
    "    x_ax.append(x_ax)\n",
    "    AL_tandem_learning(40 + i * 5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Transparency With Rational (Tandem Learning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>and</td>\n",
       "      <td>0.035263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>by</td>\n",
       "      <td>0.025491</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>with</td>\n",
       "      <td>0.025165</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>opinion</td>\n",
       "      <td>0.018606</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>its</td>\n",
       "      <td>0.015229</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>good</td>\n",
       "      <td>0.014789</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>control</td>\n",
       "      <td>0.010322</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>has</td>\n",
       "      <td>0.008529</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>there</td>\n",
       "      <td>0.008044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>sm</td>\n",
       "      <td>0.007816</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         0         1\n",
       "0      and  0.035263\n",
       "1       by  0.025491\n",
       "2     with  0.025165\n",
       "3  opinion  0.018606\n",
       "4      its  0.015229\n",
       "5     good  0.014789\n",
       "6  control  0.010322\n",
       "7      has  0.008529\n",
       "8    there  0.008044\n",
       "9       sm  0.007816"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f_we_wr[-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "t_with_rational = pd.DataFrame(stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stats=[]\n",
    "f_we_AL_no_rational = []\n",
    "def AL_no_rational(size):\n",
    "    clf = LogisticRegression(penalty='l2', C = 0.001)\n",
    "    clf.fit(X_train_vec[:size], y_train[:size])\n",
    "    \n",
    "    f_we_AL_no_rational.append(top_weighted_features(clf.coef_[0],'LR', features))\n",
    "    \n",
    "    y_predict = clf.predict(X_test_vec[:size])\n",
    "    y_proba = clf.predict_proba(X_test_vec[:size])\n",
    "    \n",
    "    a = accuracy_score(y_test[:size], y_predict)\n",
    "    preds = y_proba[:,1]\n",
    "    fpr, tpr, threshold = metrics.roc_curve(y_test[:size], preds)\n",
    "    roc_auc = metrics.auc(fpr, tpr)\n",
    "    stats.append([a,roc_auc])\n",
    "    \n",
    "    print(f\"{size}, accuracy = {a}, auc = {roc_auc}\")\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40, accuracy = 0.8, auc = 0.671875\n",
      "45, accuracy = 0.7333333333333333, auc = 0.6742424242424242\n",
      "50, accuracy = 0.68, auc = 0.5808823529411764\n",
      "55, accuracy = 0.6363636363636364, auc = 0.43\n",
      "60, accuracy = 0.6333333333333333, auc = 0.49880382775119614\n",
      "65, accuracy = 0.6153846153846154, auc = 0.46199999999999997\n",
      "70, accuracy = 0.5857142857142857, auc = 0.3994953742640875\n",
      "75, accuracy = 0.41333333333333333, auc = 0.3428981348637016\n",
      "80, accuracy = 0.5625, auc = 0.3904761904761904\n",
      "85, accuracy = 0.4117647058823529, auc = 0.35562987736900775\n",
      "90, accuracy = 0.4777777777777778, auc = 0.396833250865908\n",
      "95, accuracy = 0.45263157894736844, auc = 0.3935226264418811\n",
      "100, accuracy = 0.53, auc = 0.4109643857543017\n",
      "105, accuracy = 0.5047619047619047, auc = 0.4037763253449528\n",
      "110, accuracy = 0.4818181818181818, auc = 0.4365079365079365\n",
      "115, accuracy = 0.5217391304347826, auc = 0.42501517911353975\n",
      "120, accuracy = 0.5416666666666666, auc = 0.42097902097902096\n",
      "125, accuracy = 0.56, auc = 0.44415584415584414\n",
      "130, accuracy = 0.5692307692307692, auc = 0.4481177606177606\n",
      "135, accuracy = 0.5703703703703704, auc = 0.4433497536945813\n",
      "140, accuracy = 0.5714285714285714, auc = 0.4579166666666667\n",
      "145, accuracy = 0.5655172413793104, auc = 0.4415408439798684\n",
      "150, accuracy = 0.5666666666666667, auc = 0.45466063348416286\n",
      "155, accuracy = 0.5741935483870968, auc = 0.4560776302349336\n",
      "160, accuracy = 0.58125, auc = 0.4551436366554325\n"
     ]
    }
   ],
   "source": [
    "x_ax = []\n",
    "for i in range(25):\n",
    "    x_ax.append(40 + i * 5)\n",
    "    AL_no_rational(40 + i * 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "t_no_rational = pd.DataFrame(stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x10bed7b00>"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEICAYAAACzliQjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1122c28d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title(\"Accuracy Score across Tandem with/out rational\")\n",
    "plt.plot(x_ax, t_no_rational[0],  'r-',label=\"Tandem Without Rationale\")\n",
    "plt.plot(x_ax, t_with_rational[0],  'b-',label=\"Tandem With Rationale\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x115b07cc0>"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEICAYAAACzliQjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n0xcRqYiavohIRdT0RUQqoqYvIlKRXz96DSINx/ECAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11210f6a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title(\"AUC Score across Tandem with/out rational\")\n",
    "plt.plot(x_ax, t_no_rational[1],  'r-',label=\"Tandem Without Rationale\")\n",
    "plt.plot(x_ax, t_with_rational[1],  'b-',label=\"Tandem With Rationale\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Transparency"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "T = pd.concat([f_we_wr[-1],f_we_AL_no_rational[-1]], axis = 1, )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "T.columns =['Features(rational)', 'Weights', 'Features(no rational)', 'Weights']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Features(rational)</th>\n",
       "      <th>Weights</th>\n",
       "      <th>Features(no rational)</th>\n",
       "      <th>Weights</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>and</td>\n",
       "      <td>0.035263</td>\n",
       "      <td>on</td>\n",
       "      <td>0.011774</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>by</td>\n",
       "      <td>0.025491</td>\n",
       "      <td>to</td>\n",
       "      <td>0.008361</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>with</td>\n",
       "      <td>0.025165</td>\n",
       "      <td>get</td>\n",
       "      <td>0.008341</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>opinion</td>\n",
       "      <td>0.018606</td>\n",
       "      <td>have</td>\n",
       "      <td>0.008263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>its</td>\n",
       "      <td>0.015229</td>\n",
       "      <td>can</td>\n",
       "      <td>0.008082</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>good</td>\n",
       "      <td>0.014789</td>\n",
       "      <td>in</td>\n",
       "      <td>0.007987</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>control</td>\n",
       "      <td>0.010322</td>\n",
       "      <td>for</td>\n",
       "      <td>0.007722</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>has</td>\n",
       "      <td>0.008529</td>\n",
       "      <td>and</td>\n",
       "      <td>0.007383</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>there</td>\n",
       "      <td>0.008044</td>\n",
       "      <td>other</td>\n",
       "      <td>0.006749</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>sm</td>\n",
       "      <td>0.007816</td>\n",
       "      <td>little</td>\n",
       "      <td>0.006409</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Features(rational)   Weights Features(no rational)   Weights\n",
       "0                and  0.035263                    on  0.011774\n",
       "1                 by  0.025491                    to  0.008361\n",
       "2               with  0.025165                   get  0.008341\n",
       "3            opinion  0.018606                  have  0.008263\n",
       "4                its  0.015229                   can  0.008082\n",
       "5               good  0.014789                    in  0.007987\n",
       "6            control  0.010322                   for  0.007722\n",
       "7                has  0.008529                   and  0.007383\n",
       "8              there  0.008044                 other  0.006749\n",
       "9                 sm  0.007816                little  0.006409"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 3: Active Learning with Human annotated Rationale for 50 document"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['_id', 'asin', 'cleaned_review_text', 'helpful.0', 'helpful.1',\n",
       "       'human_labelled', 'keywords', 'labelled', 'load_to_db', 'neg_doc',\n",
       "       'overall', 'rationales', 'rationals.0', 'rationals.1', 'rationals.2',\n",
       "       'rationals.3', 'rationals.4', 'rationals.5', 'rationals.6',\n",
       "       'rationals.7', 'reviewText', 'Y'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Human labeled docs preview"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>_id</th>\n",
       "      <th>asin</th>\n",
       "      <th>cleaned_review_text</th>\n",
       "      <th>helpful.0</th>\n",
       "      <th>helpful.1</th>\n",
       "      <th>human_labelled</th>\n",
       "      <th>keywords</th>\n",
       "      <th>labelled</th>\n",
       "      <th>load_to_db</th>\n",
       "      <th>neg_doc</th>\n",
       "      <th>overall</th>\n",
       "      <th>rationales</th>\n",
       "      <th>rationals.0</th>\n",
       "      <th>rationals.1</th>\n",
       "      <th>rationals.2</th>\n",
       "      <th>rationals.3</th>\n",
       "      <th>rationals.4</th>\n",
       "      <th>rationals.5</th>\n",
       "      <th>rationals.6</th>\n",
       "      <th>rationals.7</th>\n",
       "      <th>reviewText</th>\n",
       "      <th>Y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5a1e177778b95010ea1895d2</td>\n",
       "      <td>1384719342</td>\n",
       "      <td>Not much to write about here but it does exact...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td></td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td></td>\n",
       "      <td>5.0</td>\n",
       "      <td></td>\n",
       "      <td>exactly</td>\n",
       "      <td>crisp</td>\n",
       "      <td>lowest</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>Not much to write about here, but it does exac...</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        _id        asin  \\\n",
       "0  5a1e177778b95010ea1895d2  1384719342   \n",
       "\n",
       "                                 cleaned_review_text  helpful.0  helpful.1  \\\n",
       "0  Not much to write about here but it does exact...          0          0   \n",
       "\n",
       "   human_labelled keywords  labelled  load_to_db neg_doc  overall rationales  \\\n",
       "0           False               True        True              5.0              \n",
       "\n",
       "  rationals.0 rationals.1 rationals.2 rationals.3 rationals.4 rationals.5  \\\n",
       "0     exactly       crisp      lowest                                       \n",
       "\n",
       "  rationals.6 rationals.7                                         reviewText  \\\n",
       "0                          Not much to write about here, but it does exac...   \n",
       "\n",
       "       Y  \n",
       "0  False  "
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "L = []\n",
    "y_rationa_train = []\n",
    "for row in range(train.shape[0]):\n",
    "    if train.at[row,'labelled'] == True:\n",
    "        # uncomment to see in detail\n",
    "        #print(f\"Review Text: {train.at[row,'reviewText']}\\n\\n\")\n",
    "        #print(\"****************Rationales of Text*****************\")\n",
    "        rt = ''\n",
    "        for i in range(7):\n",
    "            rt += train.at[row,f'rationals.{i}'] +','\n",
    "            #print(train.at[row,f'rationals.{i}'])\n",
    "        L.append([train.at[row,'reviewText'], rt, train.at[row,'Y']])\n",
    "        if train.at[row,'Y']:\n",
    "            y_rationa_train.append(1)\n",
    "        else:\n",
    "            y_rationa_train.append(0)\n",
    "        #print(\"***************************************************\")\n",
    "    else:\n",
    "        L.append([train.at[row,'reviewText'], '', train.at[row,'Y']])\n",
    "        if train.at[row,'Y']:\n",
    "            y_rationa_train.append(1)\n",
    "        else:\n",
    "            y_rationa_train.append(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "L_docs = pd.DataFrame(L, columns=['ReviewText', 'Rationales', 'Y'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Labelled Docs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(526, 3)"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "L_docs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ReviewText</th>\n",
       "      <th>Rationales</th>\n",
       "      <th>Y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Not much to write about here, but it does exac...</td>\n",
       "      <td>exactly,crisp,lowest,,,,,</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>The product does exactly as it should and is q...</td>\n",
       "      <td>does,affordable,,,,,,</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>The primary job of this device is to block the...</td>\n",
       "      <td>no,noticeable,reduction,secure,,,,</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Nice windscreen protects my MXL mic and preven...</td>\n",
       "      <td>prevents,pops,,,,,,</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>This pop filter is great. It looks and perform...</td>\n",
       "      <td>great,looks,performs,eliminate,pops,,,</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                          ReviewText  \\\n",
       "0  Not much to write about here, but it does exac...   \n",
       "1  The product does exactly as it should and is q...   \n",
       "2  The primary job of this device is to block the...   \n",
       "3  Nice windscreen protects my MXL mic and preven...   \n",
       "4  This pop filter is great. It looks and perform...   \n",
       "\n",
       "                               Rationales      Y  \n",
       "0               exactly,crisp,lowest,,,,,  False  \n",
       "1                   does,affordable,,,,,,  False  \n",
       "2      no,noticeable,reduction,secure,,,,  False  \n",
       "3                     prevents,pops,,,,,,  False  \n",
       "4  great,looks,performs,eliminate,pops,,,  False  "
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "L_docs.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "vectorizer = CountVectorizer(binary=True)\n",
    "X_rational_train_vec = vectorizer.fit_transform(L_docs['ReviewText'])\n",
    "X_rational_test_vec = vectorizer.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "features = pd.DataFrame(vectorizer.get_feature_names())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "features = pd.DataFrame.transpose(features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(163, 4806)"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_rational_test_vec.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(526, 4806)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_rational_train_vec.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stats = []\n",
    "f_alwor = []\n",
    "def AL_without(size):\n",
    "    \n",
    "    clf = LogisticRegression(penalty='l2', C = 0.001)\n",
    "    clf.fit(X_rational_train_vec[:size], y_rationa_train[:size])\n",
    "    \n",
    "    y_predict = clf.predict(X_rational_test_vec[:size])\n",
    "    y_proba = clf.predict_proba(X_rational_test_vec[:size])\n",
    "    \n",
    "    f_alwor.append(top_weighted_features(clf.coef_[0],'LR', features))\n",
    "    \n",
    "    \n",
    "    a = accuracy_score(y_test[:size], y_predict)\n",
    "    preds = y_proba[:,1]\n",
    "    fpr, tpr, threshold = metrics.roc_curve(y_test[:size], preds)\n",
    "    roc_auc = metrics.auc(fpr, tpr)\n",
    "    stats.append([a,roc_auc])\n",
    "    print(f\"{size}, accuracy = {a}, auc = {roc_auc}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40, accuracy = 0.8, auc = 0.6640625\n",
      "45, accuracy = 0.7333333333333333, auc = 0.6767676767676768\n",
      "50, accuracy = 0.68, auc = 0.6047794117647058\n",
      "55, accuracy = 0.6363636363636364, auc = 0.4742857142857143\n",
      "60, accuracy = 0.6333333333333333, auc = 0.5215311004784688\n",
      "65, accuracy = 0.6153846153846154, auc = 0.495\n",
      "70, accuracy = 0.5857142857142857, auc = 0.43145500420521443\n",
      "75, accuracy = 0.41333333333333333, auc = 0.3615494978479196\n",
      "80, accuracy = 0.5625, auc = 0.42666666666666664\n",
      "85, accuracy = 0.4470588235294118, auc = 0.3874024526198439\n",
      "90, accuracy = 0.4888888888888889, auc = 0.4156358238495794\n",
      "95, accuracy = 0.45263157894736844, auc = 0.40372670807453415\n",
      "100, accuracy = 0.53, auc = 0.41736694677871145\n",
      "105, accuracy = 0.5047619047619047, auc = 0.42156862745098045\n",
      "110, accuracy = 0.5, auc = 0.4642857142857143\n",
      "115, accuracy = 0.5304347826086957, auc = 0.4353369763205829\n",
      "120, accuracy = 0.55, auc = 0.4335664335664336\n",
      "125, accuracy = 0.56, auc = 0.4612987012987013\n",
      "130, accuracy = 0.5692307692307692, auc = 0.4572876447876448\n",
      "135, accuracy = 0.5703703703703704, auc = 0.4538737124944021\n",
      "140, accuracy = 0.5714285714285714, auc = 0.4654166666666667\n",
      "145, accuracy = 0.5655172413793104, auc = 0.4554781262098335\n",
      "150, accuracy = 0.5666666666666667, auc = 0.4658823529411764\n",
      "155, accuracy = 0.5741935483870968, auc = 0.46901600272386795\n",
      "160, accuracy = 0.58125, auc = 0.46637778847697003\n"
     ]
    }
   ],
   "source": [
    "for i in range(25):\n",
    "    AL_without(40 + i * 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "alwtout = pd.DataFrame(stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"Not much to write about here, but it does exactly what it's supposed to. filters out the pop sounds. now my recordings are much more crisp. it is one of the lowest prices pop filters on amazon so might as well buy it, they honestly work the same despite their pricing,\""
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "L_docs.at[0,'ReviewText']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'exactly,crisp,lowest,,,,,'"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "L_docs.at[0,'Rationales']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Rationale Addition "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "r = 1\n",
    "o = 0.01"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(526, 4806)"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_rational_train_vec = X_rational_train_vec.todense()\n",
    "X_rational_train_vec = X_rational_train_vec * o\n",
    "X_rational_train_vec.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Local Marking of rationales"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "F = {}\n",
    "for i in range(features.shape[1]):\n",
    "        F[features[i][0]] = features[i].name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "'Love' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Monster' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Placid' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Highly' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Sturdy' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Satisfied' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Replace' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'arent' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'NOT' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'WILL' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'NOT' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'EXCHANGE' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Distortion' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'Do' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n",
      "'NOT' Not Found in Vectorized Colums, \n",
      "\t\t\t\t\tNot Handled these for now-TODO\n"
     ]
    }
   ],
   "source": [
    "for row in range(X_rational_train_vec.shape[0]):\n",
    "    words = L_docs.at[row,'Rationales'].split(',')\n",
    "    for w in words:\n",
    "        if len(w) > 0:\n",
    "            try:\n",
    "                index = F[w]\n",
    "                X_rational_train_vec[row, index] = 1\n",
    "            except KeyError:\n",
    "                print(f\"'{w}' Not Found in Vectorized Colums, \\n\\t\\t\\t\\t\\tNot Handled these for now-TODO\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Test on Marking "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ReviewText</th>\n",
       "      <th>Rationales</th>\n",
       "      <th>Y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Not much to write about here, but it does exac...</td>\n",
       "      <td>exactly,crisp,lowest,,,,,</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                          ReviewText  \\\n",
       "0  Not much to write about here, but it does exac...   \n",
       "\n",
       "                  Rationales      Y  \n",
       "0  exactly,crisp,lowest,,,,,  False  "
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "L_docs.head(1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Exactly 3 words are marked in sparse matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.sum(X_rational_train_vec[0, :] == 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h2 style=\"color:green\">Section 5: Active Learning with Rationale</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stats = []\n",
    "f_we = []\n",
    "def AL_withrationale(size):\n",
    "    clf = LogisticRegression(penalty='l2', C = 0.001)\n",
    "    \n",
    "    clf.fit(X_rational_train_vec[:size], y_rationa_train[:size])\n",
    "    \n",
    "    y_predict = clf.predict(X_rational_test_vec[:size])\n",
    "    \n",
    "    y_proba = clf.predict_proba(X_rational_test_vec[:size])\n",
    "    f_we.append(top_weighted_features(clf.coef_[0],'LR', features))\n",
    "    a = accuracy_score(y_test[:size], y_predict)\n",
    "    preds = y_proba[:,1]\n",
    "    fpr, tpr, threshold = metrics.roc_curve(y_test[:size], preds)\n",
    "    roc_auc = metrics.auc(fpr, tpr)\n",
    "    stats.append([a,roc_auc])\n",
    "    print(f\"{size}, accuracy = {a}, auc = {roc_auc}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "40, accuracy = 0.8, auc = 0.49609375\n",
      "45, accuracy = 0.7333333333333333, auc = 0.4924242424242425\n",
      "50, accuracy = 0.68, auc = 0.46139705882352944\n",
      "55, accuracy = 0.6363636363636364, auc = 0.4928571428571429\n",
      "60, accuracy = 0.6333333333333333, auc = 0.49282296650717705\n",
      "65, accuracy = 0.6153846153846154, auc = 0.521\n",
      "70, accuracy = 0.5857142857142857, auc = 0.5130361648444071\n",
      "75, accuracy = 0.5466666666666666, auc = 0.47776183644189385\n",
      "80, accuracy = 0.5625, auc = 0.49269841269841275\n",
      "85, accuracy = 0.5411764705882353, auc = 0.49275362318840576\n",
      "90, accuracy = 0.5222222222222223, auc = 0.48144482929242954\n",
      "95, accuracy = 0.5157894736842106, auc = 0.4915705412599823\n",
      "100, accuracy = 0.47, auc = 0.4989995998399359\n",
      "105, accuracy = 0.47619047619047616, auc = 0.4970951343500364\n",
      "110, accuracy = 0.4727272727272727, auc = 0.49702380952380953\n",
      "115, accuracy = 0.4956521739130435, auc = 0.4820886460230723\n",
      "120, accuracy = 0.45, auc = 0.467972027972028\n",
      "125, accuracy = 0.512, auc = 0.4636363636363636\n",
      "130, accuracy = 0.5230769230769231, auc = 0.4686293436293437\n",
      "135, accuracy = 0.5481481481481482, auc = 0.45499328257948946\n",
      "140, accuracy = 0.5642857142857143, auc = 0.4510416666666666\n",
      "145, accuracy = 0.5241379310344828, auc = 0.4661246612466125\n",
      "150, accuracy = 0.5533333333333333, auc = 0.45809954751131227\n",
      "155, accuracy = 0.5741935483870968, auc = 0.4681647940074906\n",
      "160, accuracy = 0.58125, auc = 0.4660568127106403\n"
     ]
    }
   ],
   "source": [
    "for i in range(25):\n",
    "    AL_withrationale(40 + i * 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "alwr = pd.DataFrame(stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_axis=[]\n",
    "for i in range(25):\n",
    "    x_axis.append(40 + i * 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11a9f2668>"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAagAAAEYCAYAAAAJeGK1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nW2Yo8gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11a9132b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplot(121)\n",
    "plt.title(\"Accuracy Score LwR vs LwoR\")\n",
    "plt.plot(x_axis,alwr[0],'b-',label=\"LwR\")\n",
    "plt.plot(x_axis,alwtout[0], 'r-',label=\"LwoR\")\n",
    "plt.legend()\n",
    "plt.subplot(122)\n",
    "plt.title(\"AUC Score LwR vs LwoR\")\n",
    "plt.plot(x_axis,alwr[1],'b-',label=\"LwR\")\n",
    "plt.plot(x_axis,alwtout[1],'r-',label=\"LwoR\")\n",
    "plt.tight_layout()\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conclusion: Transparency Overview"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "FW = pd.concat([T, f_we[-1], f_alwor[-1]], axis=1)\n",
    "FW.columns = ['Tandem (rational, artificial labeler)', 'Weights', 'Tandem(no rational)', 'Weights', 'LwR(Human Labelled)', 'Weights', 'LwoR', 'Weights' ] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Tandem (rational, artificial labeler)</th>\n",
       "      <th>Weights</th>\n",
       "      <th>Tandem(no rational)</th>\n",
       "      <th>Weights</th>\n",
       "      <th>LwR(Human Labelled)</th>\n",
       "      <th>Weights</th>\n",
       "      <th>LwoR</th>\n",
       "      <th>Weights</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>and</td>\n",
       "      <td>0.035263</td>\n",
       "      <td>on</td>\n",
       "      <td>0.011774</td>\n",
       "      <td>great</td>\n",
       "      <td>0.002962</td>\n",
       "      <td>on</td>\n",
       "      <td>0.011758</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>by</td>\n",
       "      <td>0.025491</td>\n",
       "      <td>to</td>\n",
       "      <td>0.008361</td>\n",
       "      <td>good</td>\n",
       "      <td>0.002933</td>\n",
       "      <td>to</td>\n",
       "      <td>0.008355</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>with</td>\n",
       "      <td>0.025165</td>\n",
       "      <td>get</td>\n",
       "      <td>0.008341</td>\n",
       "      <td>quality</td>\n",
       "      <td>0.001477</td>\n",
       "      <td>get</td>\n",
       "      <td>0.008337</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>opinion</td>\n",
       "      <td>0.018606</td>\n",
       "      <td>have</td>\n",
       "      <td>0.008263</td>\n",
       "      <td>well</td>\n",
       "      <td>0.001457</td>\n",
       "      <td>have</td>\n",
       "      <td>0.008261</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>its</td>\n",
       "      <td>0.015229</td>\n",
       "      <td>can</td>\n",
       "      <td>0.008082</td>\n",
       "      <td>cables</td>\n",
       "      <td>0.001030</td>\n",
       "      <td>in</td>\n",
       "      <td>0.007907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>good</td>\n",
       "      <td>0.014789</td>\n",
       "      <td>in</td>\n",
       "      <td>0.007987</td>\n",
       "      <td>higher</td>\n",
       "      <td>0.001015</td>\n",
       "      <td>for</td>\n",
       "      <td>0.007708</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>control</td>\n",
       "      <td>0.010322</td>\n",
       "      <td>for</td>\n",
       "      <td>0.007722</td>\n",
       "      <td>horrible</td>\n",
       "      <td>0.001010</td>\n",
       "      <td>and</td>\n",
       "      <td>0.007377</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>has</td>\n",
       "      <td>0.008529</td>\n",
       "      <td>and</td>\n",
       "      <td>0.007383</td>\n",
       "      <td>doesnt</td>\n",
       "      <td>0.001005</td>\n",
       "      <td>can</td>\n",
       "      <td>0.006902</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>there</td>\n",
       "      <td>0.008044</td>\n",
       "      <td>other</td>\n",
       "      <td>0.006749</td>\n",
       "      <td>looks</td>\n",
       "      <td>0.001004</td>\n",
       "      <td>other</td>\n",
       "      <td>0.006714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>sm</td>\n",
       "      <td>0.007816</td>\n",
       "      <td>little</td>\n",
       "      <td>0.006409</td>\n",
       "      <td>recommended</td>\n",
       "      <td>0.000995</td>\n",
       "      <td>little</td>\n",
       "      <td>0.006414</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Tandem (rational, artificial labeler)   Weights Tandem(no rational)  \\\n",
       "0                                   and  0.035263                  on   \n",
       "1                                    by  0.025491                  to   \n",
       "2                                  with  0.025165                 get   \n",
       "3                               opinion  0.018606                have   \n",
       "4                                   its  0.015229                 can   \n",
       "5                                  good  0.014789                  in   \n",
       "6                               control  0.010322                 for   \n",
       "7                                   has  0.008529                 and   \n",
       "8                                 there  0.008044               other   \n",
       "9                                    sm  0.007816              little   \n",
       "\n",
       "    Weights LwR(Human Labelled)   Weights    LwoR   Weights  \n",
       "0  0.011774               great  0.002962      on  0.011758  \n",
       "1  0.008361                good  0.002933      to  0.008355  \n",
       "2  0.008341             quality  0.001477     get  0.008337  \n",
       "3  0.008263                well  0.001457    have  0.008261  \n",
       "4  0.008082              cables  0.001030      in  0.007907  \n",
       "5  0.007987              higher  0.001015     for  0.007708  \n",
       "6  0.007722            horrible  0.001010     and  0.007377  \n",
       "7  0.007383              doesnt  0.001005     can  0.006902  \n",
       "8  0.006749               looks  0.001004   other  0.006714  \n",
       "9  0.006409         recommended  0.000995  little  0.006414  "
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "FW"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Future works\n",
    "- Logistic Regression is choosed just for simplicity. But in future try other models.\n",
    "- Scale the dataset. Here I Used a small version of dataset\n",
    "- Collect Rationales more using Uncertainity Sampling\n",
    "- Extend the Web-APP UI to hold deletion.\n",
    "- Try Learning with Explanation"
   ]
  }
 ],
 "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}