LDA - Implementation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import datetime\n",
    "import numpy as np\n",
    "\n",
    "from langdetect import detect\n",
    "from tqdm import tqdm_notebook\n",
    "\n",
    "from nltk.tokenize import sent_tokenize\n",
    "from nltk.tokenize import word_tokenize\n",
    "from nltk import pos_tag\n",
    "from nltk.corpus import wordnet\n",
    "from nltk.stem.wordnet import WordNetLemmatizer\n",
    "\n",
    "from itertools import chain \n",
    "\n",
    "from gensim.models import Phrases\n",
    "from gensim import corpora\n",
    "from gensim import models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package stopwords to\n",
      "[nltk_data]     /home/deepshikha/nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "import nltk\n",
    "nltk.download('stopwords')\n",
    "\n",
    "# Using the stopwords.\n",
    "from nltk.corpus import stopwords\n",
    "\n",
    "# Initialize the stopwords\n",
    "stoplist = stopwords.words('english')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Reading Dataset and detecting language of each document"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('articles_bbc_2018_01_30.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.dropna().reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>articles</th>\n",
       "      <th>lang</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Image copyright PA/EPA Image caption Oligarch ...</td>\n",
       "      <td>en</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Husband admits killing French jogger\\r\\n\\r\\nTh...</td>\n",
       "      <td>en</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Media playback is unsupported on your device M...</td>\n",
       "      <td>en</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Manchester City's Leroy Sane is ruled out for ...</td>\n",
       "      <td>en</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Image copyright AFP Image caption Sebastien Br...</td>\n",
       "      <td>en</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            articles lang\n",
       "0  Image copyright PA/EPA Image caption Oligarch ...   en\n",
       "1  Husband admits killing French jogger\\r\\n\\r\\nTh...   en\n",
       "2  Media playback is unsupported on your device M...   en\n",
       "3  Manchester City's Leroy Sane is ruled out for ...   en\n",
       "4  Image copyright AFP Image caption Sebastien Br...   en"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "58471bcfcf6840bf859db22eceade99e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tqdm_notebook().pandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "01909faeb5ff43719e712aceeb7300d1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=308), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "data['lang'] = data.articles.progress_map(detect)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "en    257\n",
       "fa      9\n",
       "fr      7\n",
       "id      5\n",
       "uk      4\n",
       "ru      4\n",
       "hi      4\n",
       "vi      4\n",
       "ar      4\n",
       "sw      3\n",
       "tr      2\n",
       "pt      2\n",
       "es      2\n",
       "de      1\n",
       "Name: lang, dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.lang.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Filtering only English Language Articles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.loc[data.lang=='en']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Applying Sentence and Word Tokenization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1cdf25a0744946489fcbb061921dddce",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=257), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "data['sentences'] = data.articles.progress_map(sent_tokenize)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2f68753d45d74622be197ac1281ae540",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=257), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "data['tokens_sentences'] = data['sentences'].progress_map(lambda sentences: [word_tokenize(sentence) for sentence in sentences])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### POS Tagging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "67dbe881a92e4c6a97e3b3603ab5bdff",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=257), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "data['POS_tokens'] = data['tokens_sentences'].progress_map(lambda tokens_sentences: [pos_tag(tokens) for tokens in tokens_sentences])"
   ]
  },
  {
   "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>articles</th>\n",
       "      <th>lang</th>\n",
       "      <th>sentences</th>\n",
       "      <th>tokens_sentences</th>\n",
       "      <th>POS_tokens</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Image copyright PA/EPA Image caption Oligarch ...</td>\n",
       "      <td>en</td>\n",
       "      <td>[Image copyright PA/EPA Image caption Oligarch...</td>\n",
       "      <td>[[Image, copyright, PA/EPA, Image, caption, Ol...</td>\n",
       "      <td>[[(Image, NN), (copyright, NN), (PA/EPA, NNP),...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Husband admits killing French jogger\\r\\n\\r\\nTh...</td>\n",
       "      <td>en</td>\n",
       "      <td>[Husband admits killing French jogger\\r\\n\\r\\nT...</td>\n",
       "      <td>[[Husband, admits, killing, French, jogger, Th...</td>\n",
       "      <td>[[(Husband, NNP), (admits, VBZ), (killing, VBG...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Media playback is unsupported on your device M...</td>\n",
       "      <td>en</td>\n",
       "      <td>[Media playback is unsupported on your device ...</td>\n",
       "      <td>[[Media, playback, is, unsupported, on, your, ...</td>\n",
       "      <td>[[(Media, NNP), (playback, NN), (is, VBZ), (un...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Manchester City's Leroy Sane is ruled out for ...</td>\n",
       "      <td>en</td>\n",
       "      <td>[Manchester City's Leroy Sane is ruled out for...</td>\n",
       "      <td>[[Manchester, City, 's, Leroy, Sane, is, ruled...</td>\n",
       "      <td>[[(Manchester, NNP), (City, NNP), ('s, POS), (...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Image copyright AFP Image caption Sebastien Br...</td>\n",
       "      <td>en</td>\n",
       "      <td>[Image copyright AFP Image caption Sebastien B...</td>\n",
       "      <td>[[Image, copyright, AFP, Image, caption, Sebas...</td>\n",
       "      <td>[[(Image, NN), (copyright, NN), (AFP, NNP), (I...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            articles lang  \\\n",
       "0  Image copyright PA/EPA Image caption Oligarch ...   en   \n",
       "1  Husband admits killing French jogger\\r\\n\\r\\nTh...   en   \n",
       "2  Media playback is unsupported on your device M...   en   \n",
       "3  Manchester City's Leroy Sane is ruled out for ...   en   \n",
       "4  Image copyright AFP Image caption Sebastien Br...   en   \n",
       "\n",
       "                                           sentences  \\\n",
       "0  [Image copyright PA/EPA Image caption Oligarch...   \n",
       "1  [Husband admits killing French jogger\\r\\n\\r\\nT...   \n",
       "2  [Media playback is unsupported on your device ...   \n",
       "3  [Manchester City's Leroy Sane is ruled out for...   \n",
       "4  [Image copyright AFP Image caption Sebastien B...   \n",
       "\n",
       "                                    tokens_sentences  \\\n",
       "0  [[Image, copyright, PA/EPA, Image, caption, Ol...   \n",
       "1  [[Husband, admits, killing, French, jogger, Th...   \n",
       "2  [[Media, playback, is, unsupported, on, your, ...   \n",
       "3  [[Manchester, City, 's, Leroy, Sane, is, ruled...   \n",
       "4  [[Image, copyright, AFP, Image, caption, Sebas...   \n",
       "\n",
       "                                          POS_tokens  \n",
       "0  [[(Image, NN), (copyright, NN), (PA/EPA, NNP),...  \n",
       "1  [[(Husband, NNP), (admits, VBZ), (killing, VBG...  \n",
       "2  [[(Media, NNP), (playback, NN), (is, VBZ), (un...  \n",
       "3  [[(Manchester, NNP), (City, NNP), ('s, POS), (...  \n",
       "4  [[(Image, NN), (copyright, NN), (AFP, NNP), (I...  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_wordnet_pos(treebank_tag):\n",
    "\n",
    "    if treebank_tag.startswith('J'):\n",
    "        return wordnet.ADJ\n",
    "    elif treebank_tag.startswith('V'):\n",
    "        return wordnet.VERB\n",
    "    elif treebank_tag.startswith('N'):\n",
    "        return wordnet.NOUN\n",
    "    elif treebank_tag.startswith('R'):\n",
    "        return wordnet.ADV\n",
    "    else:\n",
    "        return ''\n",
    "\n",
    "lemmatizer = WordNetLemmatizer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Image', 'NN'),\n",
       " ('copyright', 'NN'),\n",
       " ('PA/EPA', 'NNP'),\n",
       " ('Image', 'NNP'),\n",
       " ('caption', 'NN'),\n",
       " ('Oligarch', 'NNP'),\n",
       " ('Roman', 'NNP'),\n",
       " ('Abramovich', 'NNP'),\n",
       " ('(', '('),\n",
       " ('l', 'NN'),\n",
       " (')', ')'),\n",
       " ('and', 'CC'),\n",
       " ('PM', 'NNP'),\n",
       " ('Dmitry', 'NNP'),\n",
       " ('Medvedev', 'NNP'),\n",
       " ('are', 'VBP'),\n",
       " ('on', 'IN'),\n",
       " ('the', 'DT'),\n",
       " ('list', 'NN'),\n",
       " ('Russian', 'NNP'),\n",
       " ('President', 'NNP'),\n",
       " ('Vladimir', 'NNP'),\n",
       " ('Putin', 'NNP'),\n",
       " ('says', 'VBZ'),\n",
       " ('a', 'DT'),\n",
       " ('list', 'NN'),\n",
       " ('of', 'IN'),\n",
       " ('officials', 'NNS'),\n",
       " ('and', 'CC'),\n",
       " ('businessmen', 'NNS'),\n",
       " ('close', 'RB'),\n",
       " ('to', 'TO'),\n",
       " ('the', 'DT'),\n",
       " ('Kremlin', 'NNP'),\n",
       " ('published', 'VBN'),\n",
       " ('by', 'IN'),\n",
       " ('the', 'DT'),\n",
       " ('US', 'NNP'),\n",
       " ('has', 'VBZ'),\n",
       " ('in', 'IN'),\n",
       " ('effect', 'NN'),\n",
       " ('targeted', 'VBN'),\n",
       " ('all', 'DT'),\n",
       " ('Russian', 'JJ'),\n",
       " ('people', 'NNS'),\n",
       " ('.', '.')]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.iloc[0]['POS_tokens'][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "595d0d910cad4c20a258b8f7369464b4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=257), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Lemmatizing each word with its POS tag, in each sentence\n",
    "data['tokens_sentences_lemmatized'] = data['POS_tokens'].progress_map(\n",
    "    lambda list_tokens_POS: [\n",
    "        [\n",
    "            lemmatizer.lemmatize(el[0], get_wordnet_pos(el[1])) \n",
    "            if get_wordnet_pos(el[1]) != '' else el[0] for el in tokens_POS\n",
    "        ] \n",
    "        for tokens_POS in list_tokens_POS\n",
    "    ]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "stopwords_verbs = ['say', 'get', 'go', 'know', 'may', 'need', 'like', 'make', 'see', 'want', 'come', 'take', 'use', 'would', 'can']\n",
    "stopwords_other = ['one', 'mr', 'bbc', 'image', 'getty', 'de', 'en', 'caption', 'also', 'copyright', 'something']\n",
    "my_stopwords = stoplist + stopwords_verbs + stopwords_other"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "data['tokens'] = data['tokens_sentences_lemmatized'].map(lambda sentences: list(chain.from_iterable(sentences)))\n",
    "data['tokens'] = data['tokens'].map(lambda tokens: [token.lower() for token in tokens if token.isalpha() \n",
    "                                                    and token.lower() not in my_stopwords and len(token)>1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['oligarch',\n",
       " 'roman',\n",
       " 'abramovich',\n",
       " 'pm',\n",
       " 'dmitry',\n",
       " 'medvedev',\n",
       " 'list',\n",
       " 'russian',\n",
       " 'president',\n",
       " 'vladimir',\n",
       " 'putin',\n",
       " 'list',\n",
       " 'official',\n",
       " 'businessmen',\n",
       " 'close',\n",
       " 'kremlin',\n",
       " 'publish',\n",
       " 'us',\n",
       " 'effect',\n",
       " 'target',\n",
       " 'russian',\n",
       " 'people',\n",
       " 'list',\n",
       " 'names',\n",
       " 'top',\n",
       " 'russians',\n",
       " 'part',\n",
       " 'sanction',\n",
       " 'law',\n",
       " 'aim']"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['tokens'].head(1).tolist()[0][:30]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokens = data['tokens'].tolist()\n",
    "bigram_model = Phrases(tokens)\n",
    "trigram_model = Phrases(bigram_model[tokens], min_count=1)\n",
    "tokens = list(trigram_model[bigram_model[tokens]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "dictionary_LDA = corpora.Dictionary(tokens)\n",
    "dictionary_LDA.filter_extremes(no_below=3)\n",
    "corpus = [dictionary_LDA.doc2bow(tok) for tok in tokens]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1.88 s, sys: 0 ns, total: 1.88 s\n",
      "Wall time: 1.95 s\n"
     ]
    }
   ],
   "source": [
    "np.random.seed(123456)\n",
    "num_topics = 20\n",
    "%time lda_model = models.LdaModel(corpus, num_topics=num_topics, \\\n",
    "                                  id2word=dictionary_LDA, \\\n",
    "                                  passes=4, alpha=[0.01]*num_topics, \\\n",
    "                                  eta=[0.01]*len(dictionary_LDA.keys()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0: 0.012*\"specie\" + 0.010*\"prey\" + 0.007*\"act\" + 0.006*\"animal\" + 0.006*\"help\" + 0.005*\"back\" + 0.005*\"find\" + 0.005*\"become\" + 0.005*\"behaviour\" + 0.005*\"number\" + 0.005*\"area\" + 0.005*\"include\" + 0.004*\"order\" + 0.004*\"kill\" + 0.004*\"approach\" + 0.004*\"however\" + 0.004*\"move\" + 0.004*\"host\" + 0.004*\"evolve\" + 0.004*\"well\"\n",
      "\n",
      "1: 0.010*\"show\" + 0.008*\"game\" + 0.007*\"another\" + 0.007*\"light\" + 0.007*\"night\" + 0.006*\"find\" + 0.006*\"could\" + 0.006*\"predator\" + 0.006*\"animal\" + 0.006*\"give\" + 0.006*\"images\" + 0.006*\"question\" + 0.006*\"however\" + 0.006*\"group\" + 0.006*\"live\" + 0.006*\"time\" + 0.005*\"transparent\" + 0.005*\"drug\" + 0.005*\"eye\" + 0.005*\"call\"\n",
      "\n",
      "2: 0.006*\"first\" + 0.006*\"people\" + 0.006*\"work\" + 0.005*\"city\" + 0.005*\"time\" + 0.005*\"year\" + 0.004*\"day\" + 0.004*\"part\" + 0.004*\"find\" + 0.004*\"could\" + 0.004*\"us\" + 0.003*\"run\" + 0.003*\"restaurant\" + 0.003*\"plan\" + 0.003*\"home\" + 0.003*\"start\" + 0.003*\"tunnel\" + 0.003*\"include\" + 0.003*\"bring\" + 0.003*\"dish\"\n",
      "\n",
      "3: 0.014*\"separatist\" + 0.011*\"people\" + 0.008*\"die\" + 0.008*\"call\" + 0.007*\"coalition\" + 0.007*\"northern_mali\" + 0.007*\"ansar_dine\" + 0.007*\"city\" + 0.006*\"force\" + 0.006*\"play\" + 0.006*\"islamist\" + 0.005*\"government\" + 0.005*\"around\" + 0.005*\"victim\" + 0.005*\"steal\" + 0.005*\"talk\" + 0.005*\"year\" + 0.005*\"prime_minister\" + 0.005*\"show\" + 0.005*\"add\"\n",
      "\n",
      "4: 0.008*\"people\" + 0.007*\"wave\" + 0.006*\"tell\" + 0.006*\"town\" + 0.006*\"water\" + 0.006*\"ocean\" + 0.005*\"benefit\" + 0.005*\"village\" + 0.005*\"world\" + 0.005*\"back\" + 0.005*\"leave\" + 0.004*\"us\" + 0.004*\"time\" + 0.004*\"fin\" + 0.004*\"work\" + 0.004*\"man\" + 0.004*\"look\" + 0.004*\"way\" + 0.004*\"first\" + 0.004*\"city\"\n",
      "\n",
      "5: 0.022*\"dutch\" + 0.011*\"issue\" + 0.010*\"border\" + 0.010*\"news\" + 0.009*\"become\" + 0.009*\"israel\" + 0.008*\"netherlands\" + 0.008*\"agency\" + 0.008*\"story\" + 0.007*\"report\" + 0.007*\"service\" + 0.007*\"un\" + 0.007*\"several\" + 0.006*\"international\" + 0.006*\"might\" + 0.006*\"part\" + 0.006*\"shy\" + 0.005*\"output\" + 0.005*\"problem\" + 0.005*\"local\"\n",
      "\n",
      "6: 0.012*\"event\" + 0.011*\"people\" + 0.009*\"president\" + 0.009*\"digital\" + 0.008*\"authority\" + 0.008*\"show\" + 0.007*\"news\" + 0.007*\"end\" + 0.007*\"election\" + 0.007*\"supporter\" + 0.007*\"correspondent\" + 0.006*\"capital\" + 0.006*\"grow\" + 0.006*\"still\" + 0.006*\"report\" + 0.006*\"well\" + 0.006*\"leave\" + 0.006*\"national\" + 0.006*\"film\" + 0.005*\"programme\"\n",
      "\n",
      "7: 0.011*\"news\" + 0.010*\"ant\" + 0.008*\"twitter\" + 0.007*\"single\" + 0.007*\"live\" + 0.006*\"name\" + 0.006*\"specie\" + 0.006*\"small\" + 0.006*\"people\" + 0.006*\"story\" + 0.005*\"welcome\" + 0.005*\"mate\" + 0.005*\"effort\" + 0.005*\"alert\" + 0.005*\"receive\" + 0.005*\"power\" + 0.005*\"millipede\" + 0.005*\"consider\" + 0.005*\"pollution\" + 0.005*\"tiny\"\n",
      "\n",
      "8: 0.012*\"us\" + 0.008*\"russia\" + 0.007*\"could\" + 0.007*\"time\" + 0.007*\"write\" + 0.006*\"president\" + 0.006*\"rocket\" + 0.006*\"call\" + 0.006*\"satellite\" + 0.006*\"trump\" + 0.005*\"many\" + 0.005*\"people\" + 0.005*\"raise\" + 0.005*\"flight\" + 0.005*\"set\" + 0.005*\"political\" + 0.005*\"tell\" + 0.004*\"try\" + 0.004*\"job\" + 0.004*\"mission\"\n",
      "\n",
      "9: 0.014*\"case\" + 0.010*\"male\" + 0.009*\"review\" + 0.009*\"prey\" + 0.008*\"find\" + 0.007*\"evidence\" + 0.007*\"kick\" + 0.007*\"message\" + 0.007*\"live\" + 0.006*\"life\" + 0.006*\"place\" + 0.006*\"study\" + 0.006*\"predator\" + 0.006*\"give\" + 0.006*\"part\" + 0.006*\"people\" + 0.006*\"number\" + 0.005*\"mate\" + 0.005*\"put\" + 0.004*\"store\"\n",
      "\n",
      "10: 0.009*\"people\" + 0.006*\"government\" + 0.006*\"uk\" + 0.006*\"think\" + 0.006*\"brexit\" + 0.005*\"tell\" + 0.005*\"time\" + 0.005*\"could\" + 0.005*\"us\" + 0.005*\"work\" + 0.004*\"even\" + 0.004*\"find\" + 0.004*\"minister\" + 0.004*\"day\" + 0.004*\"already\" + 0.004*\"help\" + 0.004*\"test\" + 0.004*\"bee\" + 0.004*\"still\" + 0.004*\"back\"\n",
      "\n",
      "11: 0.015*\"us\" + 0.014*\"stress\" + 0.011*\"healthcare\" + 0.008*\"day\" + 0.008*\"sleep\" + 0.007*\"could\" + 0.007*\"president\" + 0.007*\"test\" + 0.007*\"find\" + 0.007*\"keep\" + 0.006*\"put\" + 0.006*\"try\" + 0.006*\"images\" + 0.006*\"amazon\" + 0.006*\"provide\" + 0.006*\"team\" + 0.005*\"benefit\" + 0.005*\"expert\" + 0.005*\"body\" + 0.005*\"meeting\"\n",
      "\n",
      "12: 0.017*\"work\" + 0.013*\"people\" + 0.011*\"call\" + 0.008*\"model\" + 0.007*\"help\" + 0.007*\"find\" + 0.007*\"woman\" + 0.006*\"sleep\" + 0.006*\"spanish\" + 0.006*\"age\" + 0.006*\"radio\" + 0.006*\"happen\" + 0.006*\"report\" + 0.005*\"well\" + 0.005*\"add\" + 0.005*\"include\" + 0.005*\"accord\" + 0.005*\"police\" + 0.005*\"much\" + 0.005*\"job\"\n",
      "\n",
      "13: 0.034*\"video\" + 0.021*\"logo\" + 0.020*\"design\" + 0.015*\"platform\" + 0.014*\"tourism\" + 0.012*\"app\" + 0.012*\"animal\" + 0.012*\"country\" + 0.011*\"respect\" + 0.010*\"new\" + 0.009*\"include\" + 0.009*\"device\" + 0.009*\"call\" + 0.007*\"content\" + 0.007*\"around\" + 0.006*\"home\" + 0.006*\"others\" + 0.006*\"key\" + 0.006*\"television\" + 0.006*\"news_app\"\n",
      "\n",
      "14: 0.017*\"us\" + 0.007*\"time\" + 0.006*\"first\" + 0.006*\"america\" + 0.006*\"year\" + 0.006*\"product\" + 0.005*\"find\" + 0.005*\"amazon\" + 0.005*\"company\" + 0.005*\"still\" + 0.005*\"call\" + 0.005*\"president\" + 0.004*\"customer\" + 0.004*\"live\" + 0.004*\"store\" + 0.004*\"trump\" + 0.004*\"even\" + 0.004*\"put\" + 0.004*\"back\" + 0.004*\"light\"\n",
      "\n",
      "15: 0.014*\"list\" + 0.012*\"data\" + 0.012*\"us\" + 0.011*\"publish\" + 0.010*\"name\" + 0.010*\"could\" + 0.010*\"base\" + 0.009*\"study\" + 0.009*\"human\" + 0.007*\"russian\" + 0.007*\"security\" + 0.006*\"work\" + 0.006*\"journalist\" + 0.006*\"report\" + 0.006*\"activity\" + 0.005*\"level\" + 0.005*\"think\" + 0.005*\"system\" + 0.005*\"add\" + 0.005*\"afghanistan\"\n",
      "\n",
      "16: 0.009*\"find\" + 0.008*\"light\" + 0.008*\"us\" + 0.007*\"people\" + 0.006*\"work\" + 0.006*\"help\" + 0.005*\"day\" + 0.005*\"much\" + 0.005*\"think\" + 0.005*\"year\" + 0.005*\"images\" + 0.005*\"first\" + 0.005*\"actor\" + 0.005*\"write\" + 0.005*\"africa\" + 0.005*\"company\" + 0.004*\"right\" + 0.004*\"feel\" + 0.004*\"give\" + 0.004*\"week\"\n",
      "\n",
      "17: 0.010*\"idea\" + 0.008*\"work\" + 0.008*\"link\" + 0.007*\"news\" + 0.007*\"find\" + 0.007*\"world\" + 0.007*\"fast\" + 0.006*\"place\" + 0.006*\"information\" + 0.005*\"note\" + 0.005*\"start\" + 0.005*\"content\" + 0.005*\"chris\" + 0.005*\"audience\" + 0.005*\"human\" + 0.005*\"body\" + 0.005*\"include\" + 0.004*\"pick\" + 0.004*\"material\" + 0.004*\"journalism\"\n",
      "\n",
      "18: 0.020*\"eu\" + 0.008*\"tell\" + 0.008*\"government\" + 0.007*\"europe\" + 0.007*\"country\" + 0.006*\"visit\" + 0.006*\"include\" + 0.006*\"brexit\" + 0.006*\"work\" + 0.006*\"day\" + 0.006*\"new\" + 0.006*\"flight\" + 0.005*\"travel\" + 0.005*\"two\" + 0.005*\"deal\" + 0.005*\"vision\" + 0.005*\"couple\" + 0.005*\"reuters\" + 0.004*\"part\" + 0.004*\"route\"\n",
      "\n",
      "19: 0.026*\"video\" + 0.019*\"http\" + 0.017*\"drama\" + 0.013*\"set\" + 0.013*\"code\" + 0.011*\"always\" + 0.011*\"reference\" + 0.010*\"thank\" + 0.010*\"music\" + 0.009*\"listen\" + 0.009*\"young\" + 0.009*\"decade\" + 0.008*\"could\" + 0.008*\"live\" + 0.008*\"think\" + 0.008*\"end\" + 0.008*\"state\" + 0.008*\"love\" + 0.008*\"show\" + 0.008*\"receive\"\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for i,topic in lda_model.show_topics(formatted=True, num_topics=num_topics, num_words=20):\n",
    "    print(str(i)+\": \"+ topic)\n",
    "    print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image copyright PA/EPA Image caption Oligarch Roman Abramovich (l) and PM Dmitry Medvedev are on the list\r\n",
      "\r\n",
      "Russian President Vladimir Putin says a list of officials and businessmen close to the Kremlin published by the US has in effect targeted all Russian people.\r\n",
      "\r\n",
      "The list names 210 top Russians as part of a sanctions law aimed at punishing Moscow for meddling in the US election.\r\n",
      "\r\n",
      "However, the US stressed those named were not subject to new sanctions.\r\n",
      "\r\n",
      "Mr Putin said the list was an unfr\n"
     ]
    }
   ],
   "source": [
    "print(data.articles.loc[0][:500])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(15, 0.99830645)]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lda_model[corpus[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'This is test to check if LDA really works or not. I am using LDA along with Pyhton to detect topics out of a particular text'"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "document"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "document = '''This is test to check if LDA really works or not. I am using LDA along with Pyhton to detect topics out of a particular text'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "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 #</th>\n",
       "      <th>weight</th>\n",
       "      <th>words in topic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5</td>\n",
       "      <td>0.15</td>\n",
       "      <td>0.022*\"dutch\" + 0.011*\"issue\" + 0.010*\"border\"...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>13</td>\n",
       "      <td>0.22</td>\n",
       "      <td>0.034*\"video\" + 0.021*\"logo\" + 0.020*\"design\" ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>15</td>\n",
       "      <td>0.19</td>\n",
       "      <td>0.014*\"list\" + 0.012*\"data\" + 0.012*\"us\" + 0.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>17</td>\n",
       "      <td>0.42</td>\n",
       "      <td>0.010*\"idea\" + 0.008*\"work\" + 0.008*\"link\" + 0...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   topic #  weight                                     words in topic\n",
       "0        5    0.15  0.022*\"dutch\" + 0.011*\"issue\" + 0.010*\"border\"...\n",
       "1       13    0.22  0.034*\"video\" + 0.021*\"logo\" + 0.020*\"design\" ...\n",
       "2       15    0.19  0.014*\"list\" + 0.012*\"data\" + 0.012*\"us\" + 0.0...\n",
       "3       17    0.42  0.010*\"idea\" + 0.008*\"work\" + 0.008*\"link\" + 0..."
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokens = word_tokenize(document)\n",
    "topics = lda_model.show_topics(formatted=True, num_topics=num_topics, num_words=20)\n",
    "pd.DataFrame([(el[0], round(el[1],2), topics[el[0]][1]) for el in lda_model[dictionary_LDA.doc2bow(tokens)]], columns=['topic #', 'weight', 'words in topic'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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 #</th>\n",
       "      <th>weight</th>\n",
       "      <th>words in topic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2</td>\n",
       "      <td>0.40</td>\n",
       "      <td>0.006*\"first\" + 0.006*\"people\" + 0.006*\"work\" ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8</td>\n",
       "      <td>0.22</td>\n",
       "      <td>0.012*\"us\" + 0.008*\"russia\" + 0.007*\"could\" + ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10</td>\n",
       "      <td>0.29</td>\n",
       "      <td>0.009*\"people\" + 0.006*\"government\" + 0.006*\"u...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>17</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.010*\"idea\" + 0.008*\"work\" + 0.008*\"link\" + 0...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   topic #  weight                                     words in topic\n",
       "0        2    0.40  0.006*\"first\" + 0.006*\"people\" + 0.006*\"work\" ...\n",
       "1        8    0.22  0.012*\"us\" + 0.008*\"russia\" + 0.007*\"could\" + ...\n",
       "2       10    0.29  0.009*\"people\" + 0.006*\"government\" + 0.006*\"u...\n",
       "3       17    0.09  0.010*\"idea\" + 0.008*\"work\" + 0.008*\"link\" + 0..."
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "document = '''Eric Tucker, a 35-year-old co-founder of a marketing company in Austin, Tex., had just about 40 Twitter followers. But his recent tweet about paid protesters being bused to demonstrations against President-elect Donald J. Trump fueled a nationwide conspiracy theory — one that Mr. Trump joined in promoting. \n",
    "\n",
    "Mr. Tucker's post was shared at least 16,000 times on Twitter and more than 350,000 times on Facebook. The problem is that Mr. Tucker got it wrong. There were no such buses packed with paid protesters.\n",
    "\n",
    "But that didn't matter.\n",
    "\n",
    "While some fake news is produced purposefully by teenagers in the Balkans or entrepreneurs in the United States seeking to make money from advertising, false information can also arise from misinformed social media posts by regular people that are seized on and spread through a hyperpartisan blogosphere.\n",
    "\n",
    "Here, The New York Times deconstructs how Mr. Tucker’s now-deleted declaration on Twitter the night after the election turned into a fake-news phenomenon. It is an example of how, in an ever-connected world where speed often takes precedence over truth, an observation by a private citizen can quickly become a talking point, even as it is being proved false.'''\n",
    "tokens = word_tokenize(document)\n",
    "topics = lda_model.show_topics(formatted=True, num_topics=num_topics, num_words=20)\n",
    "pd.DataFrame([(el[0], round(el[1],2), topics[el[0]][1]) for el in lda_model[dictionary_LDA.doc2bow(tokens)]], columns=['topic #', 'weight', 'words in topic'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "topics = [lda_model[corpus[i]] for i in range(len(data))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "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>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>10</th>\n",
       "      <th>11</th>\n",
       "      <th>12</th>\n",
       "      <th>13</th>\n",
       "      <th>14</th>\n",
       "      <th>15</th>\n",
       "      <th>16</th>\n",
       "      <th>17</th>\n",
       "      <th>18</th>\n",
       "      <th>19</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0385366</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0913012</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.869287</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     0    1    2    3    4    5    6    7    8          9   10   11   12   13  \\\n",
       "0  NaN  NaN  NaN  NaN  NaN  NaN  NaN  NaN  NaN  0.0385366  NaN  NaN  NaN  NaN   \n",
       "\n",
       "    14         15   16   17        18   19  \n",
       "0  NaN  0.0913012  NaN  NaN  0.869287  NaN  "
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def topics_document_to_dataframe(topics_document, num_topics):\n",
    "    res = pd.DataFrame(columns=range(num_topics))\n",
    "    for topic_weight in topics_document:\n",
    "        res.loc[0, topic_weight[0]] = topic_weight[1]\n",
    "    return res\n",
    "\n",
    "topics_document_to_dataframe([(9, 0.03853655432967504), (15, 0.09130117862212643), (18, 0.8692868808484044)], 20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Like TF-IDF, create a matrix of topic weighting, with documents as rows and topics as columns\n",
    "document_topic = \\\n",
    "pd.concat([topics_document_to_dataframe(topics_document, num_topics=num_topics) for topics_document in topics]) \\\n",
    "  .reset_index(drop=True).fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "248    0.999237\n",
       "58     0.999117\n",
       "256    0.997293\n",
       "15     0.996790\n",
       "41     0.993013\n",
       "244    0.992777\n",
       "218    0.991440\n",
       "116    0.973609\n",
       "170    0.963455\n",
       "163    0.954756\n",
       "225    0.913619\n",
       "66     0.816265\n",
       "161    0.617414\n",
       "237    0.366396\n",
       "16     0.294009\n",
       "34     0.284695\n",
       "236    0.283284\n",
       "167    0.143202\n",
       "252    0.110725\n",
       "8      0.088885\n",
       "Name: 14, dtype: float64"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Which document are about topic 14\n",
    "document_topic.sort_values(14, ascending=False)[14].head(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image caption Brendan Cole was a professional dancer on Strictly Come Dancing for 13 years\r\n",
      "\r\n",
      "Brendan Cole has announced he will not be returning to Strictly Come Dancing.\r\n",
      "\r\n",
      "The professional dancer revealed during a TV interview on Tuesday that the decision was made by the BBC, saying he was \"in shock\".\r\n",
      "\r\n",
      "\"They made an editorial decision not to have me back on the show,\" he said on ITV's Lorraine.\r\n",
      "\r\n",
      "A spokesman for the BBC One show thanked Cole for \"being part of the show since the beginning\" and contributing to its success.\r\n",
      "\r\n",
      "'Emotional and raw'\r\n",
      "\r\n",
      "Cole said: \"I'm a little bit in shock at the moment.\r\n",
      "\r\n",
      "\"I'm quite emotional and a bit raw about it. I am very disappointed. It's an editorial decision. I will never know the ins and outs.\r\n",
      "\r\n",
      "\"I have had 15 incredible series on the show, I'm very proud of the whole show, they're a great team.\"\r\n",
      "\r\n",
      "Media playback is unsupported on your device Media caption In an interview on 16 January, Brendan Cole told 5 live he wanted to return to Stri\n"
     ]
    }
   ],
   "source": [
    "print(data.articles.loc[91][:1000])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7ff6613db748>"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x1440 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import seaborn as sns; sns.set(rc={'figure.figsize':(10,20)})\n",
    "sns.heatmap(document_topic.loc[document_topic.idxmax(axis=1).sort_values().index])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7ff6613a6390>"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set(rc={'figure.figsize':(10,5)})\n",
    "document_topic.idxmax(axis=1).value_counts().plot.bar(color='lightblue')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://cran.r-project.org/web/packages/LDAvis/vignettes/details.pdf\n",
    "# Here a short legend to explain the vis:\n",
    "# size of bubble: proportional to the proportions of the topics across the N total tokens in the corpus\n",
    "# red bars: estimated number of times a given term was generated by a given topic\n",
    "# blue bars: overall frequency of each term in the corpus\n",
    "# -- Relevance of words is computed with a parameter lambda\n",
    "# -- Lambda optimal value ~0.6 (https://nlp.stanford.edu/events/illvi2014/papers/sievert-illvi2014.pdf)\n",
    "%matplotlib inline\n",
    "import pyLDAvis\n",
    "import pyLDAvis.gensim\n",
    "vis = pyLDAvis.gensim.prepare(topic_model=lda_model, corpus=corpus, dictionary=dictionary_LDA)\n",
    "pyLDAvis.enable_notebook()\n",
    "pyLDAvis.display(vis)"
   ]
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}