twitter_trump_hostile_clf_interpretable.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load libraries\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "np.random.seed(1337) # LDiA relies on a random number generator; this keeps things constant\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "import nltk\n",
    "from nltk.tokenize.casual import casual_tokenize            # tokenize\n",
    "from nltk.tokenize.treebank import TreebankWordDetokenizer  # put it back together\n",
    "\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# the goal of this exercise is to create a model with the highest interpretability without sacrificing too much accuracy.\n",
    "# so use alphanumeric and punctuation only for tokens\n",
    "# clean out as much crap as possible from tokens\n",
    "# then do trigrams for tfidf, and inspect\n",
    "# then do ldia for topics, and inspect\n",
    "# then do classification with lda, and logistic regression for a quick check"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Step 1: Get Data\n",
    "\n",
    "The goal of this exercise is to create a model with the highest interpretability witout sacrificing too much accuracy.\n",
    "So, I will first load the data and then remove everything that isn't alphanumeric."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>tweet</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>no matter what you do for the do nothing democ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>them happy or even a little bit satisfied they...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>the most often used phrase in the lamestream m...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>just like i was right on ventilators our count...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>dems say president trump is trying to open the...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               tweet  label\n",
       "0  no matter what you do for the do nothing democ...      1\n",
       "1  them happy or even a little bit satisfied they...      1\n",
       "2  the most often used phrase in the lamestream m...      1\n",
       "3  just like i was right on ventilators our count...      0\n",
       "4  dems say president trump is trying to open the...      0"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "src_dir = r'C:\\Users\\groov\\PythonCode\\output\\trump_tweets\\\\'\n",
    "hostile_file = src_dir + 'trump_tweets_hostile_clf.csv'\n",
    "detokenizer = TreebankWordDetokenizer()\n",
    "\n",
    "clf_df = pd.read_csv(hostile_file, header=None)\n",
    "clf_df.columns = ['ts', 'tweet', 'label']\n",
    "clf_df.drop('ts', axis=1, inplace=True)\n",
    "clf_df.reset_index(drop=True, inplace=True)\n",
    "\n",
    "# wrangling\n",
    "clf_df['tweet'] = clf_df['tweet'].str.replace('&amp;', 'and')\n",
    "\n",
    "# keep only alphanum\n",
    "clf_df['tokens'] = clf_df['tweet'].apply(casual_tokenize)\n",
    "clf_df['tokens'] = clf_df['tokens'].apply(lambda tokens: [token for token in tokens if token.isalnum()])\n",
    "clf_df['tweet'] = clf_df['tokens'].apply(detokenizer.detokenize)\n",
    "\n",
    "# extract\n",
    "word_list = sorted(set(clf_df['tokens'].sum()))\n",
    "\n",
    "# clean up\n",
    "clf_df.drop('tokens', axis=1, inplace=True)\n",
    "\n",
    "clf_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's also prepare some stopwords that we can experiment with."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.corpus import stopwords\n",
    "from spacy.lang.en.stop_words import STOP_WORDS\n",
    "\n",
    "stop_words = set(stopwords.words('english'))\n",
    "combined_stop_words = set(list(stop_words) + list(STOP_WORDS))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And let's set aside the data for tfidf and bow."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "tweets = clf_df['tweet']\n",
    "labels = clf_df['label']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Step 2: Bag of Words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "\n",
    "counter = CountVectorizer(tokenizer=casual_tokenize, ngram_range=(1,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>0 such</th>\n",
       "      <th>0 such talk</th>\n",
       "      <th>0 tim</th>\n",
       "      <th>0 tim ryan</th>\n",
       "      <th>0 witnesses</th>\n",
       "      <th>00</th>\n",
       "      <th>00 a</th>\n",
       "      <th>00 a m</th>\n",
       "      <th>00 p</th>\n",
       "      <th>...</th>\n",
       "      <th>zones</th>\n",
       "      <th>zones we</th>\n",
       "      <th>zones we have</th>\n",
       "      <th>zoom</th>\n",
       "      <th>zucker</th>\n",
       "      <th>zucker will</th>\n",
       "      <th>zucker will be</th>\n",
       "      <th>zuckerberg</th>\n",
       "      <th>zuckerberg recently</th>\n",
       "      <th>zuckerberg recently stated</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 68572 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   0  0 such  0 such talk  0 tim  0 tim ryan  0 witnesses  00  00 a  00 a m  \\\n",
       "0  0       0            0      0           0            0   0     0       0   \n",
       "1  0       0            0      0           0            0   0     0       0   \n",
       "2  0       0            0      0           0            0   0     0       0   \n",
       "3  0       0            0      0           0            0   0     0       0   \n",
       "4  0       0            0      0           0            0   0     0       0   \n",
       "\n",
       "   00 p  ...  zones  zones we  zones we have  zoom  zucker  zucker will  \\\n",
       "0     0  ...      0         0              0     0       0            0   \n",
       "1     0  ...      0         0              0     0       0            0   \n",
       "2     0  ...      0         0              0     0       0            0   \n",
       "3     0  ...      0         0              0     0       0            0   \n",
       "4     0  ...      0         0              0     0       0            0   \n",
       "\n",
       "   zucker will be  zuckerberg  zuckerberg recently  zuckerberg recently stated  \n",
       "0               0           0                    0                           0  \n",
       "1               0           0                    0                           0  \n",
       "2               0           0                    0                           0  \n",
       "3               0           0                    0                           0  \n",
       "4               0           0                    0                           0  \n",
       "\n",
       "[5 rows x 68572 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bow_docs = pd.DataFrame(counter.fit_transform(clf_df['tweet'].values).toarray())\n",
    "column_nums, terms = zip(*sorted(zip(counter.vocabulary_.values(), counter.vocabulary_.keys())))\n",
    "bow_docs.columns = terms\n",
    "bow_docs.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>0 such</th>\n",
       "      <th>0 such talk</th>\n",
       "      <th>0 tim</th>\n",
       "      <th>0 tim ryan</th>\n",
       "      <th>0 witnesses</th>\n",
       "      <th>00</th>\n",
       "      <th>00 a</th>\n",
       "      <th>00 a m</th>\n",
       "      <th>00 p</th>\n",
       "      <th>...</th>\n",
       "      <th>zones we</th>\n",
       "      <th>zones we have</th>\n",
       "      <th>zoom</th>\n",
       "      <th>zucker</th>\n",
       "      <th>zucker will</th>\n",
       "      <th>zucker will be</th>\n",
       "      <th>zuckerberg</th>\n",
       "      <th>zuckerberg recently</th>\n",
       "      <th>zuckerberg recently stated</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 68573 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   0  0 such  0 such talk  0 tim  0 tim ryan  0 witnesses  00  00 a  00 a m  \\\n",
       "0  0       0            0      0           0            0   0     0       0   \n",
       "1  0       0            0      0           0            0   0     0       0   \n",
       "2  0       0            0      0           0            0   0     0       0   \n",
       "3  0       0            0      0           0            0   0     0       0   \n",
       "4  0       0            0      0           0            0   0     0       0   \n",
       "\n",
       "   00 p  ...  zones we  zones we have  zoom  zucker  zucker will  \\\n",
       "0     0  ...         0              0     0       0            0   \n",
       "1     0  ...         0              0     0       0            0   \n",
       "2     0  ...         0              0     0       0            0   \n",
       "3     0  ...         0              0     0       0            0   \n",
       "4     0  ...         0              0     0       0            0   \n",
       "\n",
       "   zucker will be  zuckerberg  zuckerberg recently  \\\n",
       "0               0           0                    0   \n",
       "1               0           0                    0   \n",
       "2               0           0                    0   \n",
       "3               0           0                    0   \n",
       "4               0           0                    0   \n",
       "\n",
       "   zuckerberg recently stated  label  \n",
       "0                           0      1  \n",
       "1                           0      1  \n",
       "2                           0      1  \n",
       "3                           0      0  \n",
       "4                           0      0  \n",
       "\n",
       "[5 rows x 68573 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's look for highest counts by class\n",
    "bow_df = bow_docs.merge(labels, left_index=True, right_index=True)\n",
    "bow_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "bow_df_agg = bow_df.groupby('label').sum().T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['the', 'and', 'to', 'a', 'is', 'of', 'in', 'that', 'i', 'for',\n",
       "       'it', 'they', 't', 'are', 'on', 'be', 'he', 'do', 'was', 's',\n",
       "       'have', 'with', 'nothing', 'will', 'democrats', 'this', 'fake',\n",
       "       'not', 'of the', 'news', 'impeachment', 'no', 'in the', 'just',\n",
       "       'all', 'you', 'now', 'fake news', 'who', 'so', 'very', 'their',\n",
       "       'president', 'has', 'do nothing', 'but', 'by', 'people', 'hoax',\n",
       "       'our', 'as', 'his', 'or', 'me', 'up', 'we', 'even', 'schiff',\n",
       "       'democrat', 'should', 'she', 'never', 'great', 'would', 'out',\n",
       "       'what', 'want', 'my', 'pelosi', 'do nothing democrats',\n",
       "       'nothing democrats', 'don', 'more', 'left', 'don t', 'house',\n",
       "       'media', 'impeachment hoax', 'to the', 'when', 'like', 'corrupt',\n",
       "       'him', 'call', 'at', 'the fake', 'nancy', 'from', 'dems', 'which',\n",
       "       'on the', 'bernie', 'and the', 'radical', 'republican', 'mike',\n",
       "       'radical left', 'were', 'party', 'if'], dtype=object)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bow_df_agg[1].sort_values(ascending=False)[0:100].index.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['the', 'and', 'to', 'of', 'a', 'in', 'will', 'is', 'for', 'our',\n",
       "       'great', 'be', 'i', 'on', 'we', 'with', 'that', 'are', 'you', 'it',\n",
       "       'have', 'will be', 'very', 'has', 'at', 'of the', 'this', 's',\n",
       "       'they', 'thank', 'by', 'he', 'all', 'thank you', 'in the', 'my',\n",
       "       'new', 'not', 'as', 'was', 'big', 'people', 'your', 'from',\n",
       "       'president', 'states', 'but', 'get', 'now', 'been', 'house',\n",
       "       'for the', 'just', 'and the', 'many', 'country', 'up', 'so',\n",
       "       'to the', 'more', 'out', 'united', 'm', 'good', 'their', 'today',\n",
       "       'a great', 'news', 'done', 'white', 'strong', 'who', 'i will',\n",
       "       'we are', 'p', 'white house', 'do', 'china', 'the united', 'job',\n",
       "       'republican', 'united states', 't', 'other', 'than',\n",
       "       'the united states', 'military', 'we will', 'p m', 'an', 'would',\n",
       "       'there', 'no', 'u', 'deal', 'doing', 'on the', 'if', 'never',\n",
       "       'with the'], dtype=object)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bow_df_agg[0].sort_values(ascending=False)[0:100].index.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Step 2: TF-IDF Vectorize\n",
    "\n",
    "Next, I'm going to use the cleaned up text and create tf-idf vectors for tokens, bigrams, and trigrams. I'll confirm this still looks good by doing basic tf-idf analysis by label. I am not going to remove stopwords, as they often contain information. For\n",
    "instance, one common slur is \"do nothing democrats\", and the word \"do\" is a stopword. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<1432x68572 sparse matrix of type '<class 'numpy.float64'>'\n",
       "\twith 137800 stored elements in Compressed Sparse Row format>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf = TfidfVectorizer(tokenizer=casual_tokenize, ngram_range=(1,3), lowercase=True, sublinear_tf=True)\n",
    "\n",
    "tfidf_docs = tfidf.fit_transform(raw_documents=tweets)\n",
    "\n",
    "tfidf_docs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>no</th>\n",
       "      <th>matter</th>\n",
       "      <th>what</th>\n",
       "      <th>you</th>\n",
       "      <th>do</th>\n",
       "      <th>for</th>\n",
       "      <th>the</th>\n",
       "      <th>nothing</th>\n",
       "      <th>democrats</th>\n",
       "      <th>how</th>\n",
       "      <th>...</th>\n",
       "      <th>number and it</th>\n",
       "      <th>and it was</th>\n",
       "      <th>it was plenty</th>\n",
       "      <th>was plenty state</th>\n",
       "      <th>plenty state should</th>\n",
       "      <th>state should have</th>\n",
       "      <th>have had them</th>\n",
       "      <th>had them in</th>\n",
       "      <th>them in stockpile</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 68573 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    no  matter  what  you   do  for  the  nothing  democrats  how  ...  \\\n",
       "0  0.0     0.0   0.0  0.0  0.0  0.0  0.0      0.0        0.0  0.0  ...   \n",
       "1  0.0     0.0   0.0  0.0  0.0  0.0  0.0      0.0        0.0  0.0  ...   \n",
       "2  0.0     0.0   0.0  0.0  0.0  0.0  0.0      0.0        0.0  0.0  ...   \n",
       "3  0.0     0.0   0.0  0.0  0.0  0.0  0.0      0.0        0.0  0.0  ...   \n",
       "4  0.0     0.0   0.0  0.0  0.0  0.0  0.0      0.0        0.0  0.0  ...   \n",
       "\n",
       "   number and it  and it was  it was plenty  was plenty state  \\\n",
       "0            0.0         0.0            0.0               0.0   \n",
       "1            0.0         0.0            0.0               0.0   \n",
       "2            0.0         0.0            0.0               0.0   \n",
       "3            0.0         0.0            0.0               0.0   \n",
       "4            0.0         0.0            0.0               0.0   \n",
       "\n",
       "   plenty state should  state should have  have had them  had them in  \\\n",
       "0                  0.0                0.0            0.0          0.0   \n",
       "1                  0.0                0.0            0.0          0.0   \n",
       "2                  0.0                0.0            0.0          0.0   \n",
       "3                  0.0                0.0            0.0          0.0   \n",
       "4                  0.0                0.0            0.0          0.0   \n",
       "\n",
       "   them in stockpile  label  \n",
       "0                0.0      1  \n",
       "1                0.0      1  \n",
       "2                0.0      1  \n",
       "3                0.0      0  \n",
       "4                0.0      0  \n",
       "\n",
       "[5 rows x 68573 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_df = pd.DataFrame(tfidf_docs.todense(), columns=tfidf.vocabulary_)\n",
    "tfidf_df = tfidf_df.merge(labels, left_index=True, right_index=True)\n",
    "\n",
    "tfidf_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>no</th>\n",
       "      <th>matter</th>\n",
       "      <th>what</th>\n",
       "      <th>you</th>\n",
       "      <th>do</th>\n",
       "      <th>for</th>\n",
       "      <th>the</th>\n",
       "      <th>nothing</th>\n",
       "      <th>democrats</th>\n",
       "      <th>how</th>\n",
       "      <th>...</th>\n",
       "      <th>that number and</th>\n",
       "      <th>number and it</th>\n",
       "      <th>and it was</th>\n",
       "      <th>it was plenty</th>\n",
       "      <th>was plenty state</th>\n",
       "      <th>plenty state should</th>\n",
       "      <th>state should have</th>\n",
       "      <th>have had them</th>\n",
       "      <th>had them in</th>\n",
       "      <th>them in stockpile</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>label</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000192</td>\n",
       "      <td>0.00012</td>\n",
       "      <td>0.00012</td>\n",
       "      <td>0.00012</td>\n",
       "      <td>0.00012</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000327</td>\n",
       "      <td>0.000159</td>\n",
       "      <td>0.000159</td>\n",
       "      <td>0.000187</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000118</td>\n",
       "      <td>0.000118</td>\n",
       "      <td>0.000118</td>\n",
       "      <td>0.000136</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00013</td>\n",
       "      <td>0.00013</td>\n",
       "      <td>0.00013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.000445</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.00047</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000253</td>\n",
       "      <td>0.000253</td>\n",
       "      <td>0.000253</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.00000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 68572 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             no   matter     what      you       do      for       the  \\\n",
       "label                                                                    \n",
       "0      0.000192  0.00012  0.00012  0.00012  0.00012  0.00000  0.000327   \n",
       "1      0.000445  0.00000  0.00000  0.00000  0.00000  0.00047  0.000000   \n",
       "\n",
       "        nothing  democrats       how  ...  that number and  number and it  \\\n",
       "label                                 ...                                   \n",
       "0      0.000159   0.000159  0.000187  ...         0.000118       0.000118   \n",
       "1      0.000000   0.000000  0.000000  ...         0.000000       0.000000   \n",
       "\n",
       "       and it was  it was plenty  was plenty state  plenty state should  \\\n",
       "label                                                                     \n",
       "0        0.000118       0.000136          0.000000             0.000000   \n",
       "1        0.000000       0.000000          0.000253             0.000253   \n",
       "\n",
       "       state should have  have had them  had them in  them in stockpile  \n",
       "label                                                                    \n",
       "0               0.000000        0.00013      0.00013            0.00013  \n",
       "1               0.000253        0.00000      0.00000            0.00000  \n",
       "\n",
       "[2 rows x 68572 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_df_agg = tfidf_df.groupby('label').mean()\n",
    "tfidf_df_agg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['that advice work', 'and safety of', 'release', 'this difficult',\n",
       "       '5', 'were perhaps', 'else it is', 'nato spending', 'than 48',\n",
       "       'can win back', 'its power it', 'd c wolves', 'report or correct',\n",
       "       'zapotosky', 'the ukraine call', 'permanent',\n",
       "       'leaking information in', 'keep us', 'many great',\n",
       "       'witnesses are up', 'the cbs', 'hoax by', 'understands the do',\n",
       "       'one billion', 'to other', 'of him', 'pelosi will go',\n",
       "       'see that new', 'with even', 'warren gets out', 'so notified',\n",
       "       'drug cartels who', 'unfortunately despite', 'trade heading',\n",
       "       'ninth circuit great', 'editorial worse than',\n",
       "       'well equipped military', 'her bernie would', 'of how badly',\n",
       "       'meeting now with', 'baker', 'wonder why', 'champ one of',\n",
       "       'and is', 'closing our northern', 'and more sinister',\n",
       "       'h o really', 'quick action', 'to 191 federal',\n",
       "       'democrats are working', 'themselves in', 'i pay',\n",
       "       'first prescription drug', 'they refuse to', 'the depressed',\n",
       "       'gee that sounds', 'a severe', 'the publicity he', 'of energy dan',\n",
       "       'economic numbers', 'poll it s', 'bill ripoffs work',\n",
       "       'with our incredible', 'depending on new', 'negative interest low',\n",
       "       'practice social', 'will be holding', 'transcripts no pressure',\n",
       "       'right reasons the', 'america hates career', 'shut out from',\n",
       "       'leading sleepy joe', 'of virginia are', 'a crime',\n",
       "       'have left and', 'nor do', 'who was', 'wants to rest',\n",
       "       'issued the china', 'openly and', 'in full retreat', 'ft high',\n",
       "       '1 the maga', 'anymore they want', 'be great', 'where are the',\n",
       "       'but publicly he', 'virtually', 'current leadership',\n",
       "       'really big rally', 's amazing', 'white house people',\n",
       "       'those of the', 'legislation that gives', 'mcconnell and kevin',\n",
       "       'in miami', 'destroy the republican', 'start of their', 'of ge',\n",
       "       'presidential primary'], dtype=object)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_df_agg.loc[1].sort_values(ascending=False)[0:100].index.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['that advice work', 'release', 'and safety of', 'else it is',\n",
       "       'her bernie would', '5', 'jail him', 'nato spending',\n",
       "       'america hates career', 'zapotosky', 'of him',\n",
       "       'presidential primary', 'this difficult', 'incompetent he made',\n",
       "       'chance a loser', 'time making the', 's has',\n",
       "       'crooked hillary fiasco', 'to other', 'd c wolves',\n",
       "       'warren gets out', 'keep us', 'are smart they', 'very unfair but',\n",
       "       'were perhaps', 'play the game', 'the publicity he',\n",
       "       'negative interest low', 'pelosi will go', 'than 48',\n",
       "       'well equipped military', 'o really', 'up version', 'made up',\n",
       "       'for some', 'and even', 'administration s', 'who was',\n",
       "       'meeting now with', 'senator rand', 'something about', 'with even',\n",
       "       'leaking information in', 'end up being', 'and is', 'one billion',\n",
       "       'long game', 'wead has written', 'see that new', 'more time doing'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_df_agg.loc[0].sort_values(ascending=False)[0:50].index.values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It doesn't seem like tf-idf analysis by label tells us much as an aggregate. I guess it is most useful for search.\n",
    "But it does provide useful features for topic modeling."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# LDiA Topic Modeling\n",
    "\n",
    "Next, we're going to use LDiA for topic modeling, as this is a smaller dataset, and LDiA allows for visibility into what\n",
    "is inside each topic. TruncatedSVD works better on larger datasets, but does not allow for inspection of the topics."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration: 1 of max_iter: 20\n",
      "iteration: 2 of max_iter: 20\n",
      "iteration: 3 of max_iter: 20\n",
      "iteration: 4 of max_iter: 20\n",
      "iteration: 5 of max_iter: 20\n",
      "iteration: 6 of max_iter: 20\n",
      "iteration: 7 of max_iter: 20\n",
      "iteration: 8 of max_iter: 20\n",
      "iteration: 9 of max_iter: 20\n",
      "iteration: 10 of max_iter: 20\n",
      "iteration: 11 of max_iter: 20\n",
      "iteration: 12 of max_iter: 20\n",
      "iteration: 13 of max_iter: 20\n",
      "iteration: 14 of max_iter: 20\n",
      "iteration: 15 of max_iter: 20\n",
      "iteration: 16 of max_iter: 20\n",
      "iteration: 17 of max_iter: 20\n",
      "iteration: 18 of max_iter: 20\n",
      "iteration: 19 of max_iter: 20\n",
      "iteration: 20 of max_iter: 20\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LatentDirichletAllocation(batch_size=128, doc_topic_prior=None,\n",
       "                          evaluate_every=-1, learning_decay=0.7,\n",
       "                          learning_method='batch', learning_offset=10.0,\n",
       "                          max_doc_update_iter=100, max_iter=20,\n",
       "                          mean_change_tol=0.001, n_components=100, n_jobs=-1,\n",
       "                          perp_tol=0.1, random_state=1337,\n",
       "                          topic_word_prior=None, total_samples=1000000.0,\n",
       "                          verbose=1)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.decomposition import LatentDirichletAllocation as LDiA\n",
    "\n",
    "ldia = LDiA(n_components=100, learning_method='batch', n_jobs=-1, random_state=1337, max_iter=20, verbose=1)\n",
    "ldia.fit(bow_docs)\n",
    "#ldia.fit(tfidf_docs)\n",
    "\n",
    "\n",
    "# try tweaking max_iter and batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "terms = list(counter.vocabulary_.keys())\n",
    "#terms = list(tfidf.vocabulary_.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>topic_0</th>\n",
       "      <th>topic_1</th>\n",
       "      <th>topic_2</th>\n",
       "      <th>topic_3</th>\n",
       "      <th>topic_4</th>\n",
       "      <th>topic_5</th>\n",
       "      <th>topic_6</th>\n",
       "      <th>topic_7</th>\n",
       "      <th>topic_8</th>\n",
       "      <th>topic_9</th>\n",
       "      <th>...</th>\n",
       "      <th>topic_90</th>\n",
       "      <th>topic_91</th>\n",
       "      <th>topic_92</th>\n",
       "      <th>topic_93</th>\n",
       "      <th>topic_94</th>\n",
       "      <th>topic_95</th>\n",
       "      <th>topic_96</th>\n",
       "      <th>topic_97</th>\n",
       "      <th>topic_98</th>\n",
       "      <th>topic_99</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>no</th>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>matter</th>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>what</th>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>you</th>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>do</th>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.01</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 100 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        topic_0  topic_1  topic_2  topic_3  topic_4  topic_5  topic_6  \\\n",
       "no         0.01     0.01     0.01     0.01     0.01     0.01     0.01   \n",
       "matter     0.01     0.01     0.01     0.01     0.01     0.01     0.01   \n",
       "what       0.01     0.01     0.01     0.01     0.01     0.01     0.01   \n",
       "you        0.01     0.01     0.01     0.01     0.01     0.01     0.01   \n",
       "do         0.01     0.01     0.01     0.01     0.01     0.01     0.01   \n",
       "\n",
       "        topic_7  topic_8  topic_9  ...  topic_90  topic_91  topic_92  \\\n",
       "no         0.01     0.01     0.01  ...      0.01      0.01      0.01   \n",
       "matter     0.01     0.01     0.01  ...      0.01      0.01      0.01   \n",
       "what       0.01     0.01     0.01  ...      0.01      0.01      0.01   \n",
       "you        0.01     0.01     0.01  ...      0.01      0.01      0.01   \n",
       "do         0.01     0.01     0.01  ...      0.01      0.01      0.01   \n",
       "\n",
       "        topic_93  topic_94  topic_95  topic_96  topic_97  topic_98  topic_99  \n",
       "no          0.01      0.01      0.01      0.01      0.01      0.01      0.01  \n",
       "matter      0.01      0.01      0.01      0.01      0.01      0.01      0.01  \n",
       "what        0.01      0.01      0.01      0.01      0.01      0.01      0.01  \n",
       "you         0.01      0.01      0.01      0.01      0.01      0.01      0.01  \n",
       "do          0.01      0.01      0.01      0.01      0.01      0.01      0.01  \n",
       "\n",
       "[5 rows x 100 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "\n",
    "components = pd.DataFrame(ldia.components_.T, index=terms)\n",
    "columns = [\"topic_{}\".format(col) for col in components.columns]\n",
    "components.columns = columns\n",
    "components.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You can see here that different words are being assigned to different topics. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(8,8))\n",
    "_= sns.heatmap(components.head(20), square=True, cbar=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's inspect the topics to see if we can find any hostile topics. Keep in mind that although the scores seem very similar \n",
    "betweet topics. there are subtle differences."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "topic_0\n",
      "\n",
      "and safety of       11.010000\n",
      "that advice work     9.636847\n",
      "release              7.673150\n",
      "this difficult       6.010000\n",
      "5                    5.974776\n",
      "d c wolves           4.869389\n",
      "were perhaps         3.010000\n",
      "nato spending        3.010000\n",
      "her bernie would     3.010000\n",
      "to other             3.010000\n",
      "Name: topic_0, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_1\n",
      "\n",
      "that advice work        46.624810\n",
      "ran the country         45.607213\n",
      "disaster that           43.010000\n",
      "rate washington         41.478595\n",
      "chance a loser          40.548089\n",
      "a mess i                40.010000\n",
      "city which is           39.831479\n",
      "incompetent he made     39.742389\n",
      "america hates career    38.270825\n",
      "nato spending           38.084239\n",
      "Name: topic_1, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_2\n",
      "\n",
      "that advice work     20.249364\n",
      "rejected their       15.010000\n",
      "addressed to my      15.010000\n",
      "not addressed to     15.010000\n",
      "5                    10.040741\n",
      "other currencies      9.010000\n",
      "has worked really     9.010000\n",
      "who has worked        9.010000\n",
      "a look                9.010000\n",
      "zapotosky             6.032003\n",
      "Name: topic_2, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_3\n",
      "\n",
      "that advice work          14.661839\n",
      "zapotosky                  8.873476\n",
      "release                    8.010000\n",
      "of him                     6.010000\n",
      "and safety of              5.757467\n",
      "her bernie would           5.010000\n",
      "a crime                    4.010000\n",
      "pelosi will go             4.010000\n",
      "crooked hillary fiasco     4.010000\n",
      "trade heading              3.010000\n",
      "Name: topic_3, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_4\n",
      "\n",
      "and safety of               10.006595\n",
      "that advice work             6.023814\n",
      "release                      4.022733\n",
      "one billion                  4.010000\n",
      "will be holding              3.069189\n",
      "superintendent of police     3.069189\n",
      "pelosi will go               3.023455\n",
      "5                            3.010000\n",
      "america hates career         3.010000\n",
      "senator rand                 2.996374\n",
      "Name: topic_4, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_5\n",
      "\n",
      "release             16.411016\n",
      "that advice work    15.414346\n",
      "and safety of       11.327773\n",
      "else it is           9.858902\n",
      "than 48              6.090480\n",
      "pelosi will go       5.414546\n",
      "her bernie would     5.192380\n",
      "nato spending        5.010000\n",
      "its power it         4.355028\n",
      "of him               4.094230\n",
      "Name: topic_5, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_6\n",
      "\n",
      "that advice work           16.868425\n",
      "5                          15.010000\n",
      "release                    14.647728\n",
      "zapotosky                   9.010000\n",
      "nato spending               8.622697\n",
      "and safety of               8.010000\n",
      "with even                   7.972850\n",
      "else it is                  7.907780\n",
      "transcripts no pressure     7.010000\n",
      "this difficult              6.010000\n",
      "Name: topic_6, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_7\n",
      "\n",
      "release                   2.01\n",
      "of him                    2.01\n",
      "tell elizabeth warren     2.01\n",
      "nato spending             2.01\n",
      "to other                  2.01\n",
      "testing is getting        1.01\n",
      "forced to leave           1.01\n",
      "be signing my             1.01\n",
      "republicans should put    1.01\n",
      "vote republican unless    1.01\n",
      "Name: topic_7, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_8\n",
      "\n",
      "that advice work      20.824814\n",
      "and safety of         11.900513\n",
      "release                9.902159\n",
      "d c wolves             6.953282\n",
      "zapotosky              5.846746\n",
      "else it is             5.525806\n",
      "5                      5.063884\n",
      "its power it           5.010000\n",
      "this difficult         5.004748\n",
      "white house people     4.044648\n",
      "Name: topic_8, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_9\n",
      "\n",
      "that advice work         28.695174\n",
      "release                  18.872827\n",
      "and safety of            11.057457\n",
      "nato spending            10.712911\n",
      "else it is                9.941148\n",
      "this difficult            9.683191\n",
      "than 48                   8.084864\n",
      "her bernie would          7.647619\n",
      "negative interest low     7.465795\n",
      "time making the           6.959544\n",
      "Name: topic_9, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_10\n",
      "\n",
      "that advice work          23.190856\n",
      "and safety of             16.205822\n",
      "nato spending             13.006700\n",
      "release                   12.956212\n",
      "warren gets out           11.838082\n",
      "leaking information in    10.886874\n",
      "this difficult             9.321741\n",
      "d c wolves                 6.515112\n",
      "the publicity he           6.010000\n",
      "were perhaps               5.997707\n",
      "Name: topic_10, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_11\n",
      "\n",
      "that advice work        28.994062\n",
      "release                 14.684381\n",
      "d c wolves              11.093305\n",
      "nato spending           10.630235\n",
      "presidential primary    10.010000\n",
      "and safety of            9.871977\n",
      "5                        9.650758\n",
      "this difficult           8.959085\n",
      "zapotosky                8.100964\n",
      "her bernie would         7.948086\n",
      "Name: topic_11, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_12\n",
      "\n",
      "that advice work    27.746652\n",
      "release             26.482351\n",
      "and safety of       12.929481\n",
      "of him              12.500955\n",
      "were perhaps        12.380335\n",
      "5                   11.009147\n",
      "else it is          10.878714\n",
      "this difficult      10.111958\n",
      "the publicity he     9.010000\n",
      "issued the china     8.654966\n",
      "Name: topic_12, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_13\n",
      "\n",
      "that advice work          51.566765\n",
      "release                   41.764196\n",
      "and safety of             22.564175\n",
      "else it is                17.191958\n",
      "her bernie would          14.997183\n",
      "of him                    13.122972\n",
      "nato spending             12.224950\n",
      "to other                   9.850468\n",
      "d c wolves                 9.302555\n",
      "well equipped military     8.886129\n",
      "Name: topic_13, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_14\n",
      "\n",
      "release                 16.248355\n",
      "and safety of            9.391356\n",
      "presidential primary     8.925609\n",
      "5                        8.694107\n",
      "zapotosky                8.010000\n",
      "pelosi will go           7.926470\n",
      "that advice work         6.664550\n",
      "d c wolves               6.290101\n",
      "time making the          4.864892\n",
      "its power it             4.793285\n",
      "Name: topic_14, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_15\n",
      "\n",
      "that advice work     21.363676\n",
      "release              14.880352\n",
      "to other              9.140483\n",
      "and safety of         7.938756\n",
      "this difficult        7.913694\n",
      "of virginia are       5.269748\n",
      "else it is            4.509830\n",
      "her bernie would      4.243426\n",
      "will ever satisfy     4.010000\n",
      "h o really            3.947041\n",
      "Name: topic_15, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_16\n",
      "\n",
      "that advice work    37.572994\n",
      "release             14.509472\n",
      "else it is          14.450936\n",
      "its power it         8.994687\n",
      "were perhaps         8.736985\n",
      "d c wolves           8.034837\n",
      "nato spending        8.010000\n",
      "than 48              7.993041\n",
      "this difficult       7.069304\n",
      "and safety of        7.044553\n",
      "Name: topic_16, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_17\n",
      "\n",
      "that advice work        8.489110\n",
      "and safety of           7.216163\n",
      "5                       6.234235\n",
      "release                 5.241356\n",
      "to other                5.014577\n",
      "have a lot              5.010000\n",
      "troops to               5.010000\n",
      "america hates career    4.060567\n",
      "cuomo wants all         4.010000\n",
      "wants all states        4.010000\n",
      "Name: topic_17, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_18\n",
      "\n",
      "that advice work          57.413693\n",
      "and safety of             30.090213\n",
      "this difficult            25.354605\n",
      "else it is                23.655938\n",
      "5                         21.205519\n",
      "release                   20.987032\n",
      "zapotosky                 13.927051\n",
      "to 191 federal            10.585837\n",
      "well equipped military    10.526655\n",
      "warren gets out           10.297939\n",
      "Name: topic_18, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_19\n",
      "\n",
      "that advice work          13.667655\n",
      "one billion                7.010000\n",
      "its power it               4.981054\n",
      "were perhaps               4.886809\n",
      "this difficult             4.010000\n",
      "leaking information in     3.950250\n",
      "keep us                    3.928202\n",
      "5                          3.570079\n",
      "else it is                 3.490342\n",
      "general the risk           3.010000\n",
      "Name: topic_19, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_20\n",
      "\n",
      "that advice work      14.540226\n",
      "d c wolves            10.472561\n",
      "release                9.038705\n",
      "and safety of          6.920373\n",
      "were perhaps           5.973476\n",
      "long game              4.036446\n",
      "the cbs                4.010000\n",
      "will the democrats     4.010000\n",
      "embarrassment all      4.010000\n",
      "all for                4.010000\n",
      "Name: topic_20, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_21\n",
      "\n",
      "that advice work          75.383850\n",
      "release                   36.509305\n",
      "else it is                31.392320\n",
      "and safety of             27.847148\n",
      "her bernie would          25.968992\n",
      "nato spending             25.289563\n",
      "d c wolves                16.038688\n",
      "well equipped military    15.692676\n",
      "zapotosky                 15.010000\n",
      "5                         13.970634\n",
      "Name: topic_21, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_22\n",
      "\n",
      "that advice work    53.182282\n",
      "and safety of       47.859194\n",
      "release             33.179684\n",
      "else it is          31.379136\n",
      "5                   31.046065\n",
      "nato spending       21.962187\n",
      "were perhaps        18.061372\n",
      "than 48             16.010000\n",
      "d c wolves          14.755328\n",
      "jail him            13.247264\n",
      "Name: topic_22, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_23\n",
      "\n",
      "that advice work        16.981970\n",
      "nato spending            6.000556\n",
      "and safety of            5.999994\n",
      "were perhaps             4.995094\n",
      "understands the do       4.010000\n",
      "this difficult           3.974857\n",
      "her bernie would         3.010000\n",
      "presidential primary     3.010000\n",
      "the ukraine call         3.010000\n",
      "to other                 3.005283\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Name: topic_23, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_24\n",
      "\n",
      "that advice work    31.849316\n",
      "release             18.766160\n",
      "nato spending       14.065951\n",
      "else it is          12.971362\n",
      "this difficult      11.032285\n",
      "and safety of       10.896697\n",
      "5                    8.054363\n",
      "than 48              7.065370\n",
      "d c wolves           7.010000\n",
      "her bernie would     6.010000\n",
      "Name: topic_24, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_25\n",
      "\n",
      "that advice work    38.232610\n",
      "and safety of       22.444536\n",
      "release             22.147541\n",
      "else it is          11.303945\n",
      "5                   10.924202\n",
      "nato spending        8.010000\n",
      "keep us              7.930395\n",
      "permanent            7.010000\n",
      "this difficult       7.010000\n",
      "were perhaps         7.010000\n",
      "Name: topic_25, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_26\n",
      "\n",
      "intolerable because the    1.981880\n",
      "do not exist               1.889036\n",
      "it away                    1.789707\n",
      "that advice work           1.640263\n",
      "the legitimate             1.010000\n",
      "his wonderful wife         1.010000\n",
      "the only reason            1.010000\n",
      "online streaming sites     1.010000\n",
      "bevin on                   1.010000\n",
      "audience is expanding      1.010000\n",
      "Name: topic_26, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_27\n",
      "\n",
      "that advice work    58.383703\n",
      "release             27.570345\n",
      "and safety of       23.853677\n",
      "else it is          18.594867\n",
      "nato spending       17.670002\n",
      "than 48             17.010000\n",
      "this difficult      16.144928\n",
      "5                   15.389602\n",
      "keep us             15.091798\n",
      "warren gets out     13.195494\n",
      "Name: topic_27, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_28\n",
      "\n",
      "that advice work    28.344989\n",
      "release             21.047866\n",
      "and safety of       17.465705\n",
      "this difficult      10.386302\n",
      "else it is          10.285723\n",
      "nato spending       10.010000\n",
      "zapotosky            9.193625\n",
      "than 48              9.124277\n",
      "were perhaps         9.114967\n",
      "of him               8.010000\n",
      "Name: topic_28, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_29\n",
      "\n",
      "that advice work        30.846484\n",
      "and safety of           26.152416\n",
      "release                 22.943629\n",
      "her bernie would        19.856919\n",
      "5                       16.002958\n",
      "else it is              15.631764\n",
      "this difficult          14.275605\n",
      "warren gets out         12.578885\n",
      "presidential primary    11.710894\n",
      "were perhaps            10.445369\n",
      "Name: topic_29, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_30\n",
      "\n",
      "that advice work    50.786042\n",
      "release             24.490717\n",
      "else it is          18.867306\n",
      "and safety of       18.823844\n",
      "nato spending       12.043481\n",
      "zapotosky           11.934685\n",
      "5                   10.248888\n",
      "her bernie would     8.816796\n",
      "of him               8.813280\n",
      "with even            6.960374\n",
      "Name: topic_30, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_31\n",
      "\n",
      "that advice work        37.297403\n",
      "release                 18.075862\n",
      "and safety of           15.109123\n",
      "this difficult          12.010000\n",
      "zapotosky                9.311125\n",
      "else it is               8.465584\n",
      "america hates career     8.168287\n",
      "than 48                  8.010000\n",
      "nato spending            7.796078\n",
      "d c wolves               6.883364\n",
      "Name: topic_31, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_32\n",
      "\n",
      "that advice work    36.208062\n",
      "and safety of       27.716800\n",
      "release             23.924261\n",
      "5                   19.460611\n",
      "her bernie would    14.743131\n",
      "else it is          14.598992\n",
      "of him              14.052838\n",
      "keep us             10.955620\n",
      "time making the     10.010000\n",
      "this difficult       9.265825\n",
      "Name: topic_32, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_33\n",
      "\n",
      "that advice work        22.277010\n",
      "release                 14.592586\n",
      "this difficult          12.959250\n",
      "zapotosky                8.010000\n",
      "her bernie would         8.010000\n",
      "presidential primary     8.010000\n",
      "else it is               7.916433\n",
      "and safety of            7.080680\n",
      "can win back             6.010000\n",
      "d c wolves               6.010000\n",
      "Name: topic_33, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_34\n",
      "\n",
      "release                 13.996734\n",
      "that advice work        13.761118\n",
      "else it is               9.789340\n",
      "5                        9.347984\n",
      "and safety of            9.000391\n",
      "to other                 8.543001\n",
      "zapotosky                8.216039\n",
      "d c wolves               8.054624\n",
      "this difficult           8.045698\n",
      "america hates career     8.010000\n",
      "Name: topic_34, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_35\n",
      "\n",
      "release                  101.100874\n",
      "presidential primary      57.371153\n",
      "that advice work          44.740770\n",
      "negative interest low     44.267947\n",
      "5                         41.321476\n",
      "else it is                40.332526\n",
      "one billion               36.105286\n",
      "this difficult            34.628832\n",
      "coming into               33.938639\n",
      "to other                  32.630576\n",
      "Name: topic_35, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_36\n",
      "\n",
      "that advice work          11.795516\n",
      "else it is                 5.974777\n",
      "and safety of              5.964926\n",
      "time making the            5.010000\n",
      "pelosi will go             4.010000\n",
      "the cbs                    4.010000\n",
      "one billion                4.002087\n",
      "d c wolves                 3.988541\n",
      "to other                   3.980298\n",
      "leaking information in     3.977861\n",
      "Name: topic_36, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_37\n",
      "\n",
      "that advice work            22.199313\n",
      "have died                   12.554112\n",
      "ever made                   11.000942\n",
      "jail him                    10.134753\n",
      "else it is                   8.606602\n",
      "brennan clapper              6.010000\n",
      "united states has            6.010000\n",
      "leader who has               6.010000\n",
      "great job for                6.010000\n",
      "congressman riggleman is     5.950144\n",
      "Name: topic_37, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_38\n",
      "\n",
      "that advice work    25.010000\n",
      "release             17.007954\n",
      "d c wolves          13.006288\n",
      "this difficult      11.003589\n",
      "5                    9.991667\n",
      "and safety of        8.010000\n",
      "else it is           6.010000\n",
      "pelosi will go       6.009367\n",
      "than 48              6.007203\n",
      "zapotosky            5.010000\n",
      "Name: topic_38, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_39\n",
      "\n",
      "that advice work          36.716650\n",
      "and safety of             16.912449\n",
      "release                   16.214214\n",
      "else it is                12.681461\n",
      "nato spending             11.880846\n",
      "this difficult            11.867002\n",
      "depending on new          11.010000\n",
      "5                         10.355118\n",
      "leaking information in     7.036866\n",
      "keep us                    6.881237\n",
      "Name: topic_39, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_40\n",
      "\n",
      "that advice work    36.870836\n",
      "release             17.233846\n",
      "and safety of       15.110509\n",
      "this difficult      13.989779\n",
      "else it is          11.071867\n",
      "5                   10.083739\n",
      "nato spending        7.136953\n",
      "than 48              6.234193\n",
      "pelosi will go       6.010000\n",
      "the cbs              6.010000\n",
      "Name: topic_40, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_41\n",
      "\n",
      "that advice work        4.941843\n",
      "jail him                3.921562\n",
      "and safety of           3.814962\n",
      "5                       3.010000\n",
      "release                 3.010000\n",
      "else it is              2.434188\n",
      "one billion             2.010000\n",
      "america hates career    2.010000\n",
      "so notified             2.010000\n",
      "they refuse to          2.010000\n",
      "Name: topic_41, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_42\n",
      "\n",
      "that advice work    48.305894\n",
      "release             40.513402\n",
      "and safety of       25.394432\n",
      "nato spending       18.211370\n",
      "else it is          17.719643\n",
      "this difficult      14.729543\n",
      "keep us             14.080123\n",
      "5                   13.757326\n",
      "its power it         9.182402\n",
      "to other             8.651928\n",
      "Name: topic_42, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_43\n",
      "\n",
      "that advice work     36.204597\n",
      "release              22.399716\n",
      "and safety of        14.473827\n",
      "5                    13.025045\n",
      "this difficult       10.090616\n",
      "else it is            9.026325\n",
      "pelosi will go        9.010000\n",
      "zapotosky             8.036812\n",
      "report or correct     8.010000\n",
      "of him                7.275931\n",
      "Name: topic_43, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_44\n",
      "\n",
      "and safety of       7.967057\n",
      "that advice work    6.577025\n",
      "d c wolves          4.963007\n",
      "release             4.961966\n",
      "5                   3.921216\n",
      "s has               3.010000\n",
      "nato spending       3.010000\n",
      "can win back        2.993264\n",
      "s report there      2.979262\n",
      "keep us             2.975161\n",
      "Name: topic_44, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_45\n",
      "\n",
      "that advice work           19.355343\n",
      "else it is                 14.650625\n",
      "release                    14.336824\n",
      "and safety of              11.407605\n",
      "her bernie would           11.405530\n",
      "one billion                10.010000\n",
      "5                           7.142815\n",
      "can win back                7.010000\n",
      "first prescription drug     7.010000\n",
      "d c wolves                  6.317945\n",
      "Name: topic_45, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_46\n",
      "\n",
      "s has                     68.244691\n",
      "are smart they            56.230516\n",
      "o really                  52.292662\n",
      "for some                  52.292662\n",
      "crooked hillary fiasco    51.569744\n",
      "very unfair but           50.718351\n",
      "play the game             49.711417\n",
      "administration s          49.549760\n",
      "her bernie would          47.150981\n",
      "of him                    45.705370\n",
      "Name: topic_46, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_47\n",
      "\n",
      "that advice work    40.272407\n",
      "and safety of       27.199391\n",
      "release             24.190981\n",
      "its power it        22.176443\n",
      "her bernie would    16.123600\n",
      "with even           14.010000\n",
      "were perhaps        13.088314\n",
      "zapotosky           13.010000\n",
      "this difficult      12.010000\n",
      "trade heading       11.010000\n",
      "Name: topic_47, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_48\n",
      "\n",
      "that advice work    27.872769\n",
      "and safety of       19.088081\n",
      "release             18.806646\n",
      "else it is          15.169076\n",
      "5                   10.069868\n",
      "her bernie would     9.865476\n",
      "d c wolves           9.204939\n",
      "nato spending        9.103294\n",
      "time making the      8.080760\n",
      "were perhaps         7.810707\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Name: topic_48, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_49\n",
      "\n",
      "and safety of       3.746741\n",
      "were perhaps        2.010000\n",
      "pent up demand      2.010000\n",
      "that advice work    2.010000\n",
      "end up being        2.010000\n",
      "the cbs             2.010000\n",
      "this difficult      2.010000\n",
      "release             2.010000\n",
      "else it is          2.010000\n",
      "every now and       2.010000\n",
      "Name: topic_49, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_50\n",
      "\n",
      "release                   14.789306\n",
      "that advice work          12.634982\n",
      "and safety of              7.010000\n",
      "5                          5.998237\n",
      "warren gets out            5.972046\n",
      "else it is                 5.928974\n",
      "well equipped military     4.987100\n",
      "to other                   4.964821\n",
      "of him                     4.010000\n",
      "margin they                4.010000\n",
      "Name: topic_50, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_51\n",
      "\n",
      "that advice work    26.952508\n",
      "and safety of       13.990326\n",
      "release             10.057204\n",
      "her bernie would     8.093764\n",
      "zapotosky            8.010000\n",
      "5                    7.795267\n",
      "of him               7.087969\n",
      "else it is           6.962772\n",
      "this difficult       6.914027\n",
      "nato spending        6.010000\n",
      "Name: topic_51, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_52\n",
      "\n",
      "that advice work    34.952606\n",
      "release             22.712720\n",
      "5                   20.043695\n",
      "and safety of       19.939264\n",
      "this difficult      12.522238\n",
      "zapotosky           12.465159\n",
      "her bernie would    12.343648\n",
      "nato spending       11.158268\n",
      "else it is          10.597912\n",
      "keep us              8.626578\n",
      "Name: topic_52, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_53\n",
      "\n",
      "that advice work    66.654451\n",
      "this difficult      25.489326\n",
      "and safety of       24.068871\n",
      "5                   22.036262\n",
      "release             18.442294\n",
      "were perhaps        14.334893\n",
      "pelosi will go      10.328015\n",
      "its power it        10.014906\n",
      "d c wolves           9.730960\n",
      "else it is           9.507131\n",
      "Name: topic_53, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_54\n",
      "\n",
      "that advice work     74.258964\n",
      "release              39.482752\n",
      "and safety of        28.172878\n",
      "5                    23.216279\n",
      "else it is           21.905044\n",
      "this difficult       18.534612\n",
      "nato spending        17.213743\n",
      "d c wolves           14.799753\n",
      "were perhaps         14.314583\n",
      "report or correct    14.160290\n",
      "Name: topic_54, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_55\n",
      "\n",
      "are fixers and          2.010000\n",
      "people i like           2.010000\n",
      "nato spending           2.010000\n",
      "radical left knowing    2.010000\n",
      "approximately 10        2.010000\n",
      "happily                 2.010000\n",
      "and the congregation    1.428515\n",
      "this difficult          1.159883\n",
      "were perhaps            1.104096\n",
      "that advice work        1.073064\n",
      "Name: topic_55, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_56\n",
      "\n",
      "release             25.933554\n",
      "that advice work    24.203224\n",
      "5                   13.964849\n",
      "and safety of       12.872655\n",
      "keep us              7.093570\n",
      "can win back         7.042810\n",
      "this difficult       6.952653\n",
      "nato spending        6.010000\n",
      "pelosi will go       5.038835\n",
      "more time doing      5.010000\n",
      "Name: topic_56, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_57\n",
      "\n",
      "and safety of         15.010000\n",
      "that advice work      14.748793\n",
      "release               11.010000\n",
      "the cbs                5.010000\n",
      "this difficult         5.010000\n",
      "very wise decision     5.010000\n",
      "her bernie would       4.905717\n",
      "else it is             4.702165\n",
      "can win back           4.010000\n",
      "nato spending          4.010000\n",
      "Name: topic_57, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_58\n",
      "\n",
      "release                 26.326527\n",
      "that advice work        26.041410\n",
      "and safety of           14.004183\n",
      "d c wolves              13.858143\n",
      "5                        9.796577\n",
      "this difficult           9.312964\n",
      "else it is               9.204500\n",
      "keep us                  8.909123\n",
      "to other                 7.928168\n",
      "america hates career     6.700729\n",
      "Name: topic_58, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_59\n",
      "\n",
      "that advice work     27.249886\n",
      "and safety of        23.095515\n",
      "this difficult       11.010276\n",
      "5                    10.647249\n",
      "else it is            9.885057\n",
      "release               8.276150\n",
      "pelosi will go        7.650379\n",
      "nato spending         7.125696\n",
      "and more sinister     6.435044\n",
      "keep us               6.353619\n",
      "Name: topic_59, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_60\n",
      "\n",
      "that advice work    15.000308\n",
      "release             12.003787\n",
      "and safety of       11.005271\n",
      "to 191 federal       6.010000\n",
      "were perhaps         6.010000\n",
      "nato spending        6.004473\n",
      "5                    5.010000\n",
      "warren gets out      5.007584\n",
      "than 48              5.007545\n",
      "else it is           5.005665\n",
      "Name: topic_60, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_61\n",
      "\n",
      "that advice work    67.362422\n",
      "this difficult      27.318069\n",
      "release             24.704322\n",
      "and safety of       24.438688\n",
      "else it is          23.959211\n",
      "nato spending       23.082511\n",
      "5                   16.977047\n",
      "keep us             12.089953\n",
      "zapotosky           12.038537\n",
      "jail him            11.346356\n",
      "Name: topic_61, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_62\n",
      "\n",
      "worst in the           19.788637\n",
      "history of debates     18.776183\n",
      "in the fbi             18.727075\n",
      "in good                12.322860\n",
      "and destabilize the     6.185585\n",
      "me with the             6.010000\n",
      "ones he was             6.010000\n",
      "release                 4.728226\n",
      "boeing                  4.298851\n",
      "more time doing         4.224981\n",
      "Name: topic_62, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_63\n",
      "\n",
      "that advice work    9.790523\n",
      "zapotosky           5.123608\n",
      "release             4.606413\n",
      "nato spending       3.205291\n",
      "d c wolves          3.003746\n",
      "the cbs             3.003515\n",
      "pelosi will go      2.993081\n",
      "else it is          2.944597\n",
      "and safety of       2.922023\n",
      "spied on            2.363296\n",
      "Name: topic_63, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_64\n",
      "\n",
      "that advice work    15.781103\n",
      "d c wolves          11.010000\n",
      "release              9.503833\n",
      "were perhaps         9.010000\n",
      "nato spending        8.875346\n",
      "else it is           6.302428\n",
      "issued the china     6.010000\n",
      "baker                5.010000\n",
      "zapotosky            4.963501\n",
      "its power it         4.010000\n",
      "Name: topic_64, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_65\n",
      "\n",
      "that advice work    75.822920\n",
      "release             37.416355\n",
      "and safety of       31.888620\n",
      "5                   16.157689\n",
      "nato spending       15.087147\n",
      "to other            14.947956\n",
      "zapotosky           14.925265\n",
      "else it is          14.395884\n",
      "her bernie would    13.982385\n",
      "warren gets out     10.942810\n",
      "Name: topic_65, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_66\n",
      "\n",
      "and safety of          15.688856\n",
      "release                15.617098\n",
      "that advice work       10.948136\n",
      "jail him                7.211056\n",
      "her bernie would        6.966280\n",
      "nato spending           5.916846\n",
      "else it is              5.759805\n",
      "exposed the             5.010000\n",
      "bachelor finale but     5.010000\n",
      "quick action            4.927966\n",
      "Name: topic_66, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_67\n",
      "\n",
      "that advice work    33.707746\n",
      "release             12.223856\n",
      "were perhaps        10.511115\n",
      "5                    9.246128\n",
      "of him               8.783634\n",
      "and safety of        8.695470\n",
      "the cbs              7.780865\n",
      "nato spending        7.458180\n",
      "warren gets out      6.703573\n",
      "else it is           6.328237\n",
      "Name: topic_67, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_68\n",
      "\n",
      "that advice work    9.01\n",
      "5                   8.01\n",
      "jail him            5.01\n",
      "else it is          5.01\n",
      "her bernie would    4.01\n",
      "rate washington     3.01\n",
      "of him              3.01\n",
      "keep us             3.01\n",
      "s has               2.01\n",
      "and safety of       2.01\n",
      "Name: topic_68, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_69\n",
      "\n",
      "pelosi will go          4.008565\n",
      "and safety of           3.987848\n",
      "that advice work        3.010000\n",
      "d c wolves              3.008082\n",
      "5                       3.006897\n",
      "nato spending           3.005021\n",
      "america hates career    3.001703\n",
      "its power it            2.010000\n",
      "permanent               2.010000\n",
      "the dnc is              2.010000\n",
      "Name: topic_69, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_70\n",
      "\n",
      "that advice work    15.472077\n",
      "and safety of       11.846010\n",
      "pelosi will go       8.993402\n",
      "5                    8.729125\n",
      "release              8.010000\n",
      "warren gets out      7.846090\n",
      "were perhaps         7.010000\n",
      "else it is           5.603551\n",
      "d c wolves           5.345729\n",
      "than 48              5.010000\n",
      "Name: topic_70, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_71\n",
      "\n",
      "that advice work    46.051710\n",
      "release             25.622103\n",
      "and safety of       17.999933\n",
      "keep us              9.010000\n",
      "nato spending        8.775254\n",
      "else it is           8.049293\n",
      "warren gets out      7.010000\n",
      "its power it         6.999244\n",
      "d c wolves           6.495122\n",
      "time making the      6.157218\n",
      "Name: topic_71, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_72\n",
      "\n",
      "that advice work      34.221360\n",
      "and safety of         17.010000\n",
      "else it is            12.091026\n",
      "the cbs                8.201144\n",
      "release                8.074366\n",
      "openly and             8.010000\n",
      "leading sleepy joe     8.010000\n",
      "its power it           7.010000\n",
      "harris poll it         7.010000\n",
      "witnesses are up       7.010000\n",
      "Name: topic_72, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_73\n",
      "\n",
      "release             21.068774\n",
      "that advice work    20.983707\n",
      "this difficult      12.807545\n",
      "5                   12.591835\n",
      "else it is          10.861984\n",
      "jail him             8.729296\n",
      "nato spending        8.699087\n",
      "and safety of        8.082377\n",
      "zapotosky            7.691777\n",
      "keep us              6.889945\n",
      "Name: topic_73, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_74\n",
      "\n",
      "release             10.549902\n",
      "d c wolves           6.010000\n",
      "see that new         5.987026\n",
      "else it is           5.815500\n",
      "that advice work     5.662812\n",
      "o really             4.988866\n",
      "for some             4.988866\n",
      "administration s     4.988647\n",
      "many great           4.987756\n",
      "drug cartels who     4.987718\n",
      "Name: topic_74, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_75\n",
      "\n",
      "one billion               4.859808\n",
      "and safety of             4.682418\n",
      "zapotosky                 4.319579\n",
      "were perhaps              4.010000\n",
      "of him                    4.010000\n",
      "practice social           4.010000\n",
      "that advice work          4.007044\n",
      "leaking information in    3.640903\n",
      "this difficult            3.345459\n",
      "dni would have            3.010000\n",
      "Name: topic_75, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_76\n",
      "\n",
      "that advice work        57.904309\n",
      "and safety of           49.085445\n",
      "release                 40.306108\n",
      "else it is              21.839267\n",
      "d c wolves              20.112883\n",
      "were perhaps            17.981921\n",
      "presidential primary    16.010000\n",
      "nato spending           11.852842\n",
      "the cbs                 10.999111\n",
      "time making the          9.733289\n",
      "Name: topic_76, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_77\n",
      "\n",
      "that advice work       6.326022\n",
      "attached to my         2.991567\n",
      "on at the              2.981828\n",
      "trade heading          2.941021\n",
      "5                      2.866144\n",
      "her bernie would       2.450181\n",
      "incompetent he made    2.010000\n",
      "nato spending          2.010000\n",
      "zapotosky              2.010000\n",
      "president has ever     2.010000\n",
      "Name: topic_77, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_78\n",
      "\n",
      "that advice work        6.312650\n",
      "were perhaps            4.259810\n",
      "america hates career    4.166310\n",
      "5                       3.513710\n",
      "zapotosky               3.150119\n",
      "wants to rest           3.010000\n",
      "and safety of           2.321325\n",
      "this difficult          2.250725\n",
      "release                 2.136842\n",
      "the cbs                 2.010000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Name: topic_78, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_79\n",
      "\n",
      "that advice work    26.855633\n",
      "release             19.841169\n",
      "and safety of       18.870139\n",
      "zapotosky           14.034836\n",
      "than 48             11.978390\n",
      "5                    9.029284\n",
      "this difficult       8.939689\n",
      "nato spending        8.932462\n",
      "d c wolves           8.045340\n",
      "of him               7.071441\n",
      "Name: topic_79, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_80\n",
      "\n",
      "that advice work    19.800686\n",
      "and safety of       11.916172\n",
      "else it is           9.782813\n",
      "nato spending        6.951396\n",
      "her bernie would     5.962618\n",
      "release              5.843545\n",
      "5                    5.695791\n",
      "meeting now with     4.987505\n",
      "of him               4.955600\n",
      "its power it         4.010000\n",
      "Name: topic_80, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_81\n",
      "\n",
      "that advice work    4.482189\n",
      "and safety of       4.479264\n",
      "nato spending       3.800537\n",
      "release             2.977314\n",
      "recovery of         2.010000\n",
      "for the speedy      2.010000\n",
      "great leaders       2.010000\n",
      "was needed good     2.010000\n",
      "can win back        2.010000\n",
      "a crime             2.010000\n",
      "Name: topic_81, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_82\n",
      "\n",
      "that advice work          37.560608\n",
      "release                   17.371168\n",
      "5                         16.084571\n",
      "nato spending             14.893608\n",
      "were perhaps              13.022930\n",
      "else it is                11.935644\n",
      "leaking information in    11.887055\n",
      "and safety of              9.917218\n",
      "keep us                    8.836455\n",
      "jail him                   7.680737\n",
      "Name: topic_82, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_83\n",
      "\n",
      "that advice work          29.909567\n",
      "else it is                15.794262\n",
      "well equipped military     9.990908\n",
      "release                    7.384881\n",
      "nato spending              6.653394\n",
      "and safety of              5.138243\n",
      "keep us                    5.010000\n",
      "federal court ruling       4.992700\n",
      "the hill                   4.992700\n",
      "zapotosky                  4.465440\n",
      "Name: topic_83, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_84\n",
      "\n",
      "that advice work          26.221207\n",
      "else it is                12.235319\n",
      "and safety of             11.721585\n",
      "well equipped military     6.430212\n",
      "jail him                   5.214925\n",
      "release                    4.121934\n",
      "nato spending              4.101762\n",
      "this difficult             4.010000\n",
      "5 p m                      4.010000\n",
      "zapotosky                  4.010000\n",
      "Name: topic_84, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_85\n",
      "\n",
      "that advice work    75.318065\n",
      "release             47.097024\n",
      "and safety of       41.180155\n",
      "else it is          27.054837\n",
      "5                   18.177607\n",
      "this difficult      16.581784\n",
      "one billion         15.161891\n",
      "zapotosky           11.766133\n",
      "to other            10.585295\n",
      "of him              10.010000\n",
      "Name: topic_85, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_86\n",
      "\n",
      "zapotosky           7.010000\n",
      "and safety of       6.010000\n",
      "release             5.033591\n",
      "that advice work    5.024647\n",
      "than 48             5.010000\n",
      "but publicly he     3.010000\n",
      "more time doing     3.010000\n",
      "can win back        3.010000\n",
      "said i              3.010000\n",
      "keep us             3.010000\n",
      "Name: topic_86, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_87\n",
      "\n",
      "that advice work        21.972996\n",
      "release                 12.453953\n",
      "and safety of            9.536931\n",
      "else it is               9.010000\n",
      "were perhaps             7.200730\n",
      "5                        6.832339\n",
      "presidential primary     5.234295\n",
      "something about          5.212373\n",
      "than 48                  5.086864\n",
      "zapotosky                5.031845\n",
      "Name: topic_87, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_88\n",
      "\n",
      "release             13.010000\n",
      "that advice work    10.740543\n",
      "and safety of        7.010000\n",
      "can win back         6.010000\n",
      "its power it         6.010000\n",
      "d c wolves           6.010000\n",
      "nato spending        5.436652\n",
      "of him               5.010000\n",
      "zapotosky            5.010000\n",
      "5                    5.010000\n",
      "Name: topic_88, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_89\n",
      "\n",
      "that advice work    38.684727\n",
      "this difficult      18.258392\n",
      "release             18.036060\n",
      "and safety of       16.348426\n",
      "than 48             13.010000\n",
      "else it is          11.117783\n",
      "time making the     11.010000\n",
      "were perhaps        10.275714\n",
      "witnesses are up     9.010000\n",
      "who was              9.010000\n",
      "Name: topic_89, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_90\n",
      "\n",
      "that advice work    17.087377\n",
      "5                    9.876807\n",
      "and safety of        8.810815\n",
      "release              8.334962\n",
      "else it is           7.407590\n",
      "than 48              7.010719\n",
      "nato spending        6.423444\n",
      "its power it         6.010000\n",
      "this difficult       5.666407\n",
      "with even            5.312138\n",
      "Name: topic_90, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_91\n",
      "\n",
      "that advice work    14.01\n",
      "zapotosky           10.01\n",
      "and safety of        8.01\n",
      "her bernie would     8.01\n",
      "of him               6.01\n",
      "d c wolves           5.01\n",
      "nato spending        5.01\n",
      "one billion          5.01\n",
      "so notified          4.01\n",
      "its power it         4.01\n",
      "Name: topic_91, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_92\n",
      "\n",
      "that advice work    13.209828\n",
      "release             10.126452\n",
      "5                    5.010000\n",
      "were perhaps         4.207011\n",
      "this difficult       4.047062\n",
      "its power it         4.010000\n",
      "trade heading        3.062606\n",
      "nato spending        3.052842\n",
      "and safety of        3.021261\n",
      "permanent            3.010000\n",
      "Name: topic_92, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_93\n",
      "\n",
      "that advice work    26.308167\n",
      "and safety of       11.597250\n",
      "5                    9.574512\n",
      "d c wolves           7.752215\n",
      "release              7.010000\n",
      "zapotosky            7.010000\n",
      "were perhaps         6.999925\n",
      "this difficult       6.684307\n",
      "else it is           6.628975\n",
      "pelosi will go       5.988082\n",
      "Name: topic_93, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_94\n",
      "\n",
      "release             5.918564\n",
      "and safety of       5.261534\n",
      "keep us             3.666381\n",
      "than 48             3.010000\n",
      "of him              3.010000\n",
      "zapotosky           2.909824\n",
      "of energy dan       2.010000\n",
      "issued the china    2.010000\n",
      "5                   2.010000\n",
      "to say about        2.010000\n",
      "Name: topic_94, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_95\n",
      "\n",
      "that advice work    63.873736\n",
      "release             35.986968\n",
      "else it is          23.727697\n",
      "and safety of       20.540891\n",
      "were perhaps        16.968440\n",
      "5                   13.431462\n",
      "nato spending       12.922127\n",
      "zapotosky           11.054846\n",
      "than 48             10.906769\n",
      "keep us             10.852907\n",
      "Name: topic_95, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_96\n",
      "\n",
      "5                         3.351548\n",
      "release                   3.010000\n",
      "not your                  2.010000\n",
      "in terms of               2.010000\n",
      "that they would           2.010000\n",
      "than 48                   2.010000\n",
      "the onslaught of          2.010000\n",
      "signed giant trade        2.010000\n",
      "that advice work          1.241964\n",
      "insubordinate reported    1.010000\n",
      "Name: topic_96, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_97\n",
      "\n",
      "that advice work        107.185974\n",
      "release                  46.858722\n",
      "and safety of            36.506898\n",
      "else it is               29.842095\n",
      "this difficult           22.651116\n",
      "nato spending            22.614921\n",
      "were perhaps             22.348122\n",
      "5                        21.750302\n",
      "to other                 19.666964\n",
      "america hates career     19.294018\n",
      "Name: topic_97, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_98\n",
      "\n",
      "that advice work    28.008327\n",
      "release             16.241354\n",
      "and safety of       15.037397\n",
      "else it is          12.010000\n",
      "this difficult       9.228219\n",
      "5                    9.025022\n",
      "zapotosky            7.214326\n",
      "issued the china     7.010000\n",
      "with even            7.010000\n",
      "permanent            6.058700\n",
      "Name: topic_98, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n",
      "topic_99\n",
      "\n",
      "release                 11.868749\n",
      "that advice work         8.696923\n",
      "else it is               5.819914\n",
      "america hates career     5.806486\n",
      "d c wolves               5.010000\n",
      "nato spending            4.975255\n",
      "zapotosky                4.869881\n",
      "5                        4.506290\n",
      "one billion              4.010000\n",
      "pelosi will go           4.003463\n",
      "Name: topic_99, dtype: float64\n",
      "\n",
      "---------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "for col in components.columns:\n",
    "    print(col)\n",
    "    print()\n",
    "    output = components[col].sort_values(ascending=False).head(10)\n",
    "    print(output)\n",
    "    print()\n",
    "    print('---------------------------------------------------------')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So, that's some interesting visibility into the topics. That's one very nice thing about LDiA. So, now, onto prediction.\n",
    "\n",
    "But first, let's make the topic vectors for use in prediction."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>topic_0</th>\n",
       "      <th>topic_1</th>\n",
       "      <th>topic_2</th>\n",
       "      <th>topic_3</th>\n",
       "      <th>topic_4</th>\n",
       "      <th>topic_5</th>\n",
       "      <th>topic_6</th>\n",
       "      <th>topic_7</th>\n",
       "      <th>topic_8</th>\n",
       "      <th>topic_9</th>\n",
       "      <th>...</th>\n",
       "      <th>topic_90</th>\n",
       "      <th>topic_91</th>\n",
       "      <th>topic_92</th>\n",
       "      <th>topic_93</th>\n",
       "      <th>topic_94</th>\n",
       "      <th>topic_95</th>\n",
       "      <th>topic_96</th>\n",
       "      <th>topic_97</th>\n",
       "      <th>topic_98</th>\n",
       "      <th>topic_99</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.000065</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.993311</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>0.000068</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.151815</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>0.000066</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 100 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    topic_0   topic_1   topic_2   topic_3   topic_4   topic_5   topic_6  \\\n",
       "0  0.000060  0.000060  0.000060  0.000060  0.000060  0.000060  0.000060   \n",
       "1  0.000065  0.000065  0.000065  0.000065  0.000065  0.000065  0.000065   \n",
       "2  0.000068  0.000068  0.000068  0.000068  0.000068  0.000068  0.993311   \n",
       "3  0.000072  0.000072  0.000072  0.000072  0.000072  0.000072  0.000072   \n",
       "4  0.000066  0.000066  0.000066  0.000066  0.000066  0.000066  0.000066   \n",
       "\n",
       "    topic_7   topic_8   topic_9  ...  topic_90  topic_91  topic_92  topic_93  \\\n",
       "0  0.000060  0.000060  0.000060  ...  0.000060  0.000060  0.000060  0.000060   \n",
       "1  0.000065  0.000065  0.000065  ...  0.000065  0.000065  0.000065  0.000065   \n",
       "2  0.000068  0.000068  0.000068  ...  0.000068  0.000068  0.000068  0.000068   \n",
       "3  0.000072  0.000072  0.000072  ...  0.000072  0.000072  0.000072  0.000072   \n",
       "4  0.000066  0.000066  0.151815  ...  0.000066  0.000066  0.000066  0.000066   \n",
       "\n",
       "   topic_94  topic_95  topic_96  topic_97  topic_98  topic_99  \n",
       "0  0.000060  0.000060  0.000060  0.000060  0.000060  0.000060  \n",
       "1  0.000065  0.000065  0.000065  0.000065  0.000065  0.000065  \n",
       "2  0.000068  0.000068  0.000068  0.000068  0.000068  0.000068  \n",
       "3  0.000072  0.000072  0.000072  0.000072  0.000072  0.000072  \n",
       "4  0.000066  0.000066  0.000066  0.000066  0.000066  0.000066  \n",
       "\n",
       "[5 rows x 100 columns]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "topic_vectors = ldia.transform(bow_docs)\n",
    "#topic_vectors = ldia.transform(tfidf_docs)\n",
    "\n",
    "columns = ['topic_{}'.format(i) for i in range(ldia.n_components)]\n",
    "topic_vectors = pd.DataFrame(topic_vectors, columns=columns)\n",
    "topic_vectors.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hostile Classification\n",
    "\n",
    "This part is a work in progress. Get ideas from this guide and possibly re-read the chapter from the book.\n",
    "https://www.machinelearningplus.com/nlp/topic-modeling-python-sklearn-examples/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1432, 100), (1432,))"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = topic_vectors\n",
    "y = labels\n",
    "\n",
    "X.shape, y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((716, 100), (716, 100), 716, 716)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1337, test_size=0.5)\n",
    "X_train.shape, X_test.shape, len(y_train), len(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.6997206703910615, 0.5879888268156425)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA\n",
    "\n",
    "lda = LDA(n_components=1, solver='lsqr', shrinkage=0.4)\n",
    "lda.fit(X_train, y_train)\n",
    "lda.score(X_train, y_train), lda.score(X_test, y_test)\n",
    "\n",
    "# LatentDirichletAllocation topic modeling\n",
    "\n",
    "# (0.6731843575418994, 0.5279329608938548) - bow w/o stopwords\n",
    "# (0.6871508379888268, 0.5572625698324022) - bow w/ all characters w/o stopwords\n",
    "# (0.6885474860335196, 0.5502793296089385) - bow w/ all characters and all stopwords\n",
    "# (0.6871508379888268, 0.5502793296089385) - more ldia runs... so doesn't improve. moving on\n",
    "\n",
    "# (0.6620111731843575, 0.5544692737430168) - tfidf w/o stopwords\n",
    "# (0.6564245810055865, 0.5614525139664804) - tfidf w/ all characters w/o stopwords\n",
    "# (0.7262569832402235, 0.6075418994413407) - tfidf w/ all characters and all stopwords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "bow_1 = [0.673, 0.528]\n",
    "bow_2 = [0.687, 0.557]\n",
    "bow_3 = [0.689, 0.550]\n",
    "\n",
    "tfidf_1 = [0.662, 0.554]\n",
    "tfidf_2 = [0.656, 0.561]\n",
    "tfidf_3 = [0.736, 0.608]\n",
    "\n",
    "res_df = pd.DataFrame({'1':bow_1, '2':bow_2, '3':bow_3, '4':tfidf_1, '5':tfidf_2, '6':tfidf_3})\n",
    "res_df.columns = ['bow_alpha_text_no_stopwords', 'bow_all_text_no_stopwords', 'bow_complete',\n",
    "                 'tfidf_alpha_text_no_stopwords', 'tfidf_all_text_no_stopwords', 'tfidf_complete']\n",
    "res_df = res_df.T\n",
    "res_df.columns = ['Train', 'Test']\n",
    "_= res_df.plot(figsize=(15,5), marker='o', title='LDiA Classification Scores w/ BoW and TF-IDF', subplots=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Topic Modeling with TruncatedSVD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import TruncatedSVD\n",
    "\n",
    "svd = TruncatedSVD(n_components=100, n_iter=100, random_state=1337)\n",
    "#svd_topic_vectors = svd.fit_transform(bow_docs)\n",
    "svd_topic_vectors = svd.fit_transform(tfidf_docs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1432, 100), (1432,))"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = svd_topic_vectors\n",
    "y = labels\n",
    "\n",
    "X.shape, y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((716, 100), (716, 100), 716, 716)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1337, test_size=0.5)\n",
    "X_train.shape, X_test.shape, len(y_train), len(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.9050279329608939, 0.8784916201117319)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lda = LDA(n_components=1, solver='lsqr', shrinkage=0.4)\n",
    "lda.fit(X_train, y_train)\n",
    "lda.score(X_train, y_train), lda.score(X_test, y_test)\n",
    "\n",
    "# truncatedSVD topic modeling\n",
    "# (0.8729050279329609, 0.8337988826815642) - bow w/ all characters and all stopwords\n",
    "# (0.909217877094972, 0.8980446927374302) - tfidf w/ all characters and all stopwords\n",
    "# (0.9050279329608939, 0.8784916201117319) - tfidf alphanum and all stopwords - simple and interpretable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "topic_df = pd.DataFrame(X).merge(labels, left_index=True, right_index=True)\n",
    "topic_df_0 = topic_df[topic_df['label']==0].iloc[:,0:100]\n",
    "topic_df_1 = topic_df[topic_df['label']==1].iloc[:,0:100]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(15,5))\n",
    "_= sns.heatmap(topic_df_0.head(20), cbar=False, square=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(15,5))\n",
    "_= sns.heatmap(topic_df_1.head(20), cbar=False, square=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 792x792 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(11,11))\n",
    "_= sns.heatmap(topic_df.head(100), cbar=False, square=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}