word2vec_got.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Word2Vec e Game of Thrones \n",
    "Neste notebook, faremos uma análise dos roteiros da série Game of Thrones utilizando Word2Vec. Este dataset pode ser encontrado [aqui](https://www.kaggle.com/anderfj/game-of-thrones-series-scripts-breakdowns) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Breve explicação\n",
    "A suposição feita pelo word2vec é a de que duas palavras que compartilham contextos semelhantes, também irão compartilhar uma representação vetorial semelhante do modelo. Por exemplo, se palavras como \"cachorrinho\" e \"fofo\" forem frequentemente usadas em situações semelhantes, logo, essas mesmas palavras também terão vetores similares em sua representação segundo o word2vec. \n",
    "\n",
    "É através desta suposição que o word2vec consegue encontrar relações entre as palavras em um dataset, computar a similaridade entre elas e usar a sua representação em vetores para essas palavras como input para outras aplicações, como classificação de texto ou clustering. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Importando bibliotecas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import re \n",
    "from time import time\n",
    "from collections import defaultdict \n",
    "import spacy \n",
    "import logging\n",
    "logging.basicConfig(format=\"%(levelname)s - %(asctime)s: %(message)s\", datefmt= '%H:%M:%S', level=logging.INFO)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pré-processamentos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Carregando o dataset \n",
    "df = pd.read_csv('got_scripts_breakdown.csv', sep = ';')\n",
    "df.drop('Column1', axis=1, inplace=True)"
   ]
  },
  {
   "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>Season</th>\n",
       "      <th>Episode</th>\n",
       "      <th>Sentence</th>\n",
       "      <th>Name</th>\n",
       "      <th>N_serie</th>\n",
       "      <th>N_Season</th>\n",
       "      <th>Emision Date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Season 1</td>\n",
       "      <td>winter is coming</td>\n",
       "      <td>What d’you expect? They’re savages One lot ste...</td>\n",
       "      <td>waymar royce</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17/04/2011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Season 1</td>\n",
       "      <td>winter is coming</td>\n",
       "      <td>I’ve never seen wildlings do a thing like this...</td>\n",
       "      <td>will</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17/04/2011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Season 1</td>\n",
       "      <td>winter is coming</td>\n",
       "      <td>How close did you get?</td>\n",
       "      <td>waymar royce</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17/04/2011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Season 1</td>\n",
       "      <td>winter is coming</td>\n",
       "      <td>Close as any man would</td>\n",
       "      <td>will</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17/04/2011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Season 1</td>\n",
       "      <td>winter is coming</td>\n",
       "      <td>We should head back to the wall</td>\n",
       "      <td>gared</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17/04/2011</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Season           Episode  \\\n",
       "0  Season 1  winter is coming   \n",
       "1  Season 1  winter is coming   \n",
       "2  Season 1  winter is coming   \n",
       "3  Season 1  winter is coming   \n",
       "4  Season 1  winter is coming   \n",
       "\n",
       "                                            Sentence          Name  N_serie  \\\n",
       "0  What d’you expect? They’re savages One lot ste...  waymar royce        1   \n",
       "1  I’ve never seen wildlings do a thing like this...          will        1   \n",
       "2                             How close did you get?  waymar royce        1   \n",
       "3                             Close as any man would          will        1   \n",
       "4                    We should head back to the wall         gared        1   \n",
       "\n",
       "   N_Season Emision Date  \n",
       "0         1   17/04/2011  \n",
       "1         1   17/04/2011  \n",
       "2         1   17/04/2011  \n",
       "3         1   17/04/2011  \n",
       "4         1   17/04/2011  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Leitura das primeiras linhas do dataset\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(22510, 7)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Criando funções para limpar o texto"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "nlp = spacy.load('en_core_web_sm')\n",
    "\n",
    "def limpar_texto(doc):\n",
    "    \"\"\"\n",
    "    Lematiza e retira stopwords.\n",
    "    O documento precisa ser um objeto spacy\n",
    "    \"\"\"\n",
    "    \n",
    "    texto = [token.lemma_ for token in doc if not token.is_stop]\n",
    "    \n",
    "    if len(texto)>2:\n",
    "        return \" \".join(texto)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#retirando caracteres que não são letras e passando todos eles para sua forma minúscula.\n",
    "brief_cleaning = (re.sub(\"[^A-Za-z']+\", ' ', str(row)).lower() for row in df['Sentence']) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para aplicar os pré-processamentos, estaremos utilizando a pipeline da biblioteca spaCy. Você pode ler mais sobre isso [aqui](https://spacy.io/api/language#pipe) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tempo para limpar tudo: 0.83 mins\n"
     ]
    }
   ],
   "source": [
    "t = time()\n",
    "\n",
    "#Aplicando as funções \n",
    "txt = [limpar_texto(doc) for doc in nlp.pipe(brief_cleaning, batch_size=5000)]\n",
    "\n",
    "print('Tempo para limpar tudo: {} mins'.format(round((time() - t) / 60, 2))) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_limpo = pd.DataFrame({'texto_limpo': txt})\n",
    "df_limpo = df_limpo.dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>texto_limpo</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>d expect savage lot steal goat lot know rip piece</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ve see wildling thing like ve see thing like life</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>order track wildling track win t trouble</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>don t think will ask die horse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>s good thing child want run away south run awa...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                         texto_limpo\n",
       "0  d expect savage lot steal goat lot know rip piece\n",
       "1  ve see wildling thing like ve see thing like life\n",
       "6           order track wildling track win t trouble\n",
       "7                     don t think will ask die horse\n",
       "9  s good thing child want run away south run awa..."
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_limpo.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(22510, 7)\n",
      "(13761, 1)\n"
     ]
    }
   ],
   "source": [
    "print(df.shape)\n",
    "print(df_limpo.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "No processo de limpar o nosso texto, retiramos muitas linhas do nosso dataset original que eram compostas apenas por stopwords. Por isso, geramos um novo dataframe apenas com o texto limpo."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Bigramas\n",
    "Vamos utilizar o pacote [Gensim Phrases](https://radimrehurek.com/gensim/models/phrases.html) para automaticamente detectar bigramas. Com isso, conseguimos captar expressões como \"jon_snow\" e \"tyrion_lannister\". "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gensim.models.phrases import Phrases, Phraser"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Como o `Phrases()` recebe como input uma lista de lista de palavras, vamos fazer esse passo primeiro: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "sent = [row.split() for row in df_limpo['texto_limpo']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Criando bigramas relevantes a partir da lista de frases:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO - 14:16:04: collecting all words and their counts\n",
      "INFO - 14:16:04: PROGRESS: at sentence #0, processed 0 words and 0 word types\n",
      "INFO - 14:16:05: PROGRESS: at sentence #10000, processed 77822 words and 54321 word types\n",
      "INFO - 14:16:05: collected 72644 word types from a corpus of 110250 words (unigram + bigrams) and 13761 sentences\n",
      "INFO - 14:16:05: using 72644 counts as vocab in Phrases<0 vocab, min_count=30, threshold=10.0, max_vocab_size=40000000>\n",
      "INFO - 14:16:05: source_vocab length 72644\n",
      "INFO - 14:16:05: Phraser built with 30 phrasegrams\n"
     ]
    }
   ],
   "source": [
    "phrases = Phrases(sent, min_count=30, progress_per=10000)\n",
    "bigram = Phraser(phrases)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['you', 'know', 'nothing', 'jon_snow']\n"
     ]
    }
   ],
   "source": [
    "print(bigram[\"you know nothing jon snow\".split()])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Como podemos observar, o nome \"Jon Snow\" foi reconhecido como um bigrama (indicado pelo `_` )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentences = bigram[sent]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Word2Vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import multiprocessing\n",
    "\n",
    "from gensim.models import Word2Vec"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Os parâmetros\n",
    "Há uma lista bem extensa com os parâmetros que podemos usar no [word2vec](https://radimrehurek.com/gensim/models/word2vec.html). Aqui, vamos apenas ajustar alguns, e, para os demais, usaremos a configuração padrão.\n",
    "\n",
    "- `size`: representa aqui a dimensão do espaço vetorial. Esse parâmetro costuma ficar em um intervalo de 50 a 300. Como o nosso dataset é relativamente pequeno, vamos usar **80**\n",
    "- `min_count`: contagem mínima que uma palavra precisa aparecer para ser treinada. Como já tínhamos colocado isso no pré-processamento, aqui vamos usar **2** também.\n",
    "- `window`: distância entre a palavra atual e a que queremos prever - **2**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "w2v_model = Word2Vec(min_count=2, size=80, window=2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Construindo a tabela com o vocabulário\n",
    "Word2Vec requer a construção de uma tabela com o vocabulário (simplesmente seleciona todas as palavras e filtra as que são únicas. Além de fazer algumas contagens com elas)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO - 14:16:06: collecting all words and their counts\n",
      "INFO - 14:16:06: PROGRESS: at sentence #0, processed 0 words, keeping 0 word types\n",
      "INFO - 14:16:06: PROGRESS: at sentence #10000, processed 76471 words, keeping 5688 word types\n",
      "INFO - 14:16:06: collected 6675 word types from a corpus of 108063 raw words and 13761 sentences\n",
      "INFO - 14:16:06: Loading a fresh vocabulary\n",
      "INFO - 14:16:06: effective_min_count=2 retains 4109 unique words (61% of original 6675, drops 2566)\n",
      "INFO - 14:16:06: effective_min_count=2 leaves 105497 word corpus (97% of original 108063, drops 2566)\n",
      "INFO - 14:16:06: deleting the raw counts dictionary of 6675 items\n",
      "INFO - 14:16:06: sample=0.001 downsamples 62 most-common words\n",
      "INFO - 14:16:06: downsampling leaves estimated 92177 word corpus (87.4% of prior 105497)\n",
      "INFO - 14:16:06: estimated required memory for 4109 words and 80 dimensions: 4684260 bytes\n",
      "INFO - 14:16:06: resetting layer weights\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time to build vocab: 0.03 mins\n"
     ]
    }
   ],
   "source": [
    "t = time()\n",
    "\n",
    "w2v_model.build_vocab(sentences, progress_per=10000) \n",
    "\n",
    "print('Time to build vocab: {} mins'.format(round((time() - t) / 60, 2)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Treinando o modelo \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO - 14:16:07: training model with 3 workers on 4109 vocabulary and 80 features, using sg=0 hs=0 sample=0.001 negative=5 window=2\n",
      "INFO - 14:16:08: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:08: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:08: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:08: EPOCH - 1 : training on 108063 raw words (92190 effective words) took 0.5s, 179242 effective words/s\n",
      "INFO - 14:16:08: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:08: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:08: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:08: EPOCH - 2 : training on 108063 raw words (92100 effective words) took 0.5s, 178739 effective words/s\n",
      "INFO - 14:16:09: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:09: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:09: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:09: EPOCH - 3 : training on 108063 raw words (92226 effective words) took 0.6s, 167428 effective words/s\n",
      "INFO - 14:16:09: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:09: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:09: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:09: EPOCH - 4 : training on 108063 raw words (92218 effective words) took 0.5s, 180014 effective words/s\n",
      "INFO - 14:16:10: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:10: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:10: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:10: EPOCH - 5 : training on 108063 raw words (92204 effective words) took 0.5s, 182542 effective words/s\n",
      "INFO - 14:16:10: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:10: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:10: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:10: EPOCH - 6 : training on 108063 raw words (92170 effective words) took 0.5s, 182820 effective words/s\n",
      "INFO - 14:16:11: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:11: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:11: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:11: EPOCH - 7 : training on 108063 raw words (92291 effective words) took 0.6s, 161680 effective words/s\n",
      "INFO - 14:16:12: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:12: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:12: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:12: EPOCH - 8 : training on 108063 raw words (92352 effective words) took 0.5s, 184380 effective words/s\n",
      "INFO - 14:16:12: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:12: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:12: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:12: EPOCH - 9 : training on 108063 raw words (92130 effective words) took 0.5s, 183357 effective words/s\n",
      "INFO - 14:16:13: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:13: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:13: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:13: EPOCH - 10 : training on 108063 raw words (92291 effective words) took 0.6s, 163451 effective words/s\n",
      "INFO - 14:16:13: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:13: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:13: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:13: EPOCH - 11 : training on 108063 raw words (92310 effective words) took 0.5s, 185035 effective words/s\n",
      "INFO - 14:16:14: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:14: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:14: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:14: EPOCH - 12 : training on 108063 raw words (92165 effective words) took 0.5s, 179139 effective words/s\n",
      "INFO - 14:16:14: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:14: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:14: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:14: EPOCH - 13 : training on 108063 raw words (92229 effective words) took 0.5s, 170137 effective words/s\n",
      "INFO - 14:16:15: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:15: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:15: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:15: EPOCH - 14 : training on 108063 raw words (92032 effective words) took 0.5s, 177470 effective words/s\n",
      "INFO - 14:16:15: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:15: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:15: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:15: EPOCH - 15 : training on 108063 raw words (92234 effective words) took 0.5s, 187229 effective words/s\n",
      "INFO - 14:16:16: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:16: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:16: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:16: EPOCH - 16 : training on 108063 raw words (92163 effective words) took 0.5s, 189423 effective words/s\n",
      "INFO - 14:16:16: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:16: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:16: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:16: EPOCH - 17 : training on 108063 raw words (92113 effective words) took 0.6s, 164774 effective words/s\n",
      "INFO - 14:16:17: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:17: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:17: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:17: EPOCH - 18 : training on 108063 raw words (92157 effective words) took 0.5s, 187496 effective words/s\n",
      "INFO - 14:16:17: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:17: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:17: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:17: EPOCH - 19 : training on 108063 raw words (92259 effective words) took 0.5s, 187263 effective words/s\n",
      "INFO - 14:16:18: worker thread finished; awaiting finish of 2 more threads\n",
      "INFO - 14:16:18: worker thread finished; awaiting finish of 1 more threads\n",
      "INFO - 14:16:18: worker thread finished; awaiting finish of 0 more threads\n",
      "INFO - 14:16:18: EPOCH - 20 : training on 108063 raw words (92192 effective words) took 0.6s, 166064 effective words/s\n",
      "INFO - 14:16:18: training on a 2161260 raw words (1844026 effective words) took 10.6s, 173842 effective words/s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time to train the model: 0.18 mins\n"
     ]
    }
   ],
   "source": [
    "t = time()\n",
    "\n",
    "w2v_model.train(sentences, total_examples=w2v_model.corpus_count, epochs=20, report_delay=1)\n",
    "\n",
    "print('Time to train the model: {} mins'.format(round((time() - t) / 60, 2)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Explorando o modelo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `most_similar`\n",
    "\n",
    "Vamos explorar quais são as palavras mais similares de acordo com alguns elementos da série:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO - 14:16:18: precomputing L2-norms of word weight vectors\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('ramsay', 0.8904601335525513),\n",
       " ('snow', 0.8364033102989197),\n",
       " ('yield', 0.8292680978775024),\n",
       " ('bolton', 0.8284211158752441),\n",
       " ('survive', 0.8095954656600952),\n",
       " ('ironborn', 0.8039368987083435),\n",
       " ('capital', 0.8007180690765381),\n",
       " ('robb', 0.7996101975440979),\n",
       " ('castle_black', 0.7994767427444458),\n",
       " ('riverrun', 0.7992897033691406)]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.most_similar(positive=[\"winterfell\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('defend', 0.8826789855957031),\n",
       " ('slaughter', 0.834825873374939),\n",
       " ('abandon', 0.8246427774429321),\n",
       " ('join', 0.8137688040733337),\n",
       " ('protect', 0.8084516525268555),\n",
       " ('flay', 0.807414710521698),\n",
       " ('return', 0.7912930250167847),\n",
       " ('kingdom', 0.7865685224533081),\n",
       " ('entire', 0.7796071767807007),\n",
       " ('survive', 0.7769665718078613)]"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.most_similar(positive=['night_watch'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('stormborn', 0.9505660533905029),\n",
       " ('targaryen', 0.9080790281295776),\n",
       " ('aegon', 0.829805850982666),\n",
       " ('aerys', 0.7972592115402222),\n",
       " ('rhaegar', 0.7906438112258911),\n",
       " ('rhaenyra', 0.7883229851722717),\n",
       " ('unburnt', 0.7823358178138733),\n",
       " ('unhurt', 0.7700403928756714),\n",
       " ('queen', 0.7662924528121948),\n",
       " ('visenya', 0.7571664452552795)]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.most_similar('daenerys')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Parece que a longa lista de nomes da Daenerys foi bem captada pelo nosso modelo rs.\n",
    "\n",
    "Vamos ver quem é o candidato mais discutido quando o assunto é o Trono de Ferro segundo o nosso modelo:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('throne', 0.9030369520187378),\n",
       " ('proclaim', 0.8675816059112549),\n",
       " ('prosperity', 0.8613318800926208),\n",
       " ('stannis_baratheon', 0.8555351495742798),\n",
       " ('troublesome', 0.8463995456695557),\n",
       " ('stead', 0.845191240310669),\n",
       " ('steffon', 0.843341052532196),\n",
       " ('royal', 0.8421745300292969),\n",
       " ('restoration', 0.8398019075393677),\n",
       " ('side', 0.8384577035903931)]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.most_similar(positive=[\"iron_throne\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Dentre outras palavras, parece que o personagem principal inserido nesse contexto é o Stannis Baratheon.\n",
    "\n",
    "Além disso, também podemos usar essa função para fazer analogias. Qual será que é a palavra que está para \"king\", assim como \"queen\" está para \"cersei\"? "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "def analogy(x1, x2, y1):\n",
    "    result = w2v_model.wv.most_similar(positive = [y1, x2], negative=[x1])\n",
    "    return result [0][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'tywin_lannister'"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "analogy('queen', 'cersei', 'king')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `similarity`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Agora, vamos ver quão similares são duas palavras diferentes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.34574005"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.similarity('jon', 'king')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6299548"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.similarity('joffrey', 'king')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `doesnt_match`\n",
    "Esta função pede para que o nosso modelo encontre o a palavra que não pertence à lista. Aqui, vamos tentar utilizá-la para reconhecer o membro que não pertence àquela família:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/julia/opt/anaconda3/lib/python3.7/site-packages/gensim/models/keyedvectors.py:877: FutureWarning: arrays to stack must be passed as a \"sequence\" type such as list or tuple. Support for non-sequence iterables such as generators is deprecated as of NumPy 1.16 and will raise an error in the future.\n",
      "  vectors = vstack(self.word_vec(word, use_norm=True) for word in used_words).astype(REAL)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'joffrey'"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.doesnt_match(['jon', 'arya', 'joffrey', 'sansa'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'daenerys'"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w2v_model.wv.doesnt_match(['cersei', 'jaime', 'daenerys', 'tyrion'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualizações "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    " \n",
    "import seaborn as sns\n",
    "sns.set_style(\"darkgrid\")\n",
    "\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.manifold import TSNE"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* A intenção do notebook é a de explicar como aplicar o word2vec e não como o t-SNE funciona. Se você quiser saber mais sobre isso, você pode começar por [aqui](https://medium.com/@luckylwk/visualising-high-dimensional-datasets-using-pca-and-t-sne-in-python-8ef87e7915b)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nosso objetivo nesta parte é plotar 80 dimensões de vetores em gráficos bidimensionais, e, com isso, observar se conseguimos encontrar padrões interessantes. \n",
    "\n",
    "Para fazer as visualizações mais interessantes, vamos observar os relacionamentos entre a palavra da query (em vermelho), as palavras correspondentes mais semelhantes ao modelo (em azul), e outras palavras do vocabulário (em verde). \n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def tsnescatterplot(model, word, list_names):\n",
    "\n",
    "    arrays = np.empty((0, 80), dtype='f')\n",
    "    word_labels = [word]\n",
    "    color_list  = ['red']\n",
    "\n",
    "    # adiciona o vetor da palavra da query\n",
    "    arrays = np.append(arrays, model.wv.__getitem__([word]), axis=0)\n",
    "    \n",
    "    # gera lista com as palavras mais similares \n",
    "    close_words = model.wv.most_similar([word])\n",
    "    \n",
    "    # adiciona os vetores para cada uma das palavras mais próximas no array \n",
    "    for wrd_score in close_words:\n",
    "        wrd_vector = model.wv.__getitem__([wrd_score[0]])\n",
    "        word_labels.append(wrd_score[0])\n",
    "        color_list.append('blue')\n",
    "        arrays = np.append(arrays, wrd_vector, axis=0)\n",
    "    \n",
    "    # adiciona os vetores para cada uma das palavras em list_names no array \n",
    "    for wrd in list_names:\n",
    "        wrd_vector = model.wv.__getitem__([wrd])\n",
    "        word_labels.append(wrd)\n",
    "        color_list.append('green')\n",
    "        arrays = np.append(arrays, wrd_vector, axis=0)\n",
    "        \n",
    "    # Reduz a dimensionalidade de 80 para 10 dimensões com PCA\n",
    "    reduc = PCA(n_components=10).fit_transform(arrays)\n",
    "    \n",
    "    # Acha as coordenadas do t-SNE para 2 dimensões \n",
    "    np.set_printoptions(suppress=True)\n",
    "    \n",
    "    Y = TSNE(n_components=2, random_state=0, perplexity=15).fit_transform(reduc)\n",
    "    \n",
    "    # Arruma tudo para o plot\n",
    "    df = pd.DataFrame({'x': [x for x in Y[:, 0]],\n",
    "                       'y': [y for y in Y[:, 1]],\n",
    "                       'words': word_labels,\n",
    "                       'color': color_list})\n",
    "    \n",
    "    fig, _ = plt.subplots()\n",
    "    fig.set_size_inches(9, 9)\n",
    "    \n",
    "    # Plot básico\n",
    "    p1 = sns.regplot(data=df,\n",
    "                     x=\"x\",\n",
    "                     y=\"y\",\n",
    "                     fit_reg=False,\n",
    "                     marker=\"o\",\n",
    "                     scatter_kws={'s': 40,\n",
    "                                  'facecolors': df['color']\n",
    "                                 }\n",
    "                    )\n",
    "    \n",
    "    # Adiciona anotações uma por uma com um loop\n",
    "    for line in range(0, df.shape[0]):\n",
    "         p1.text(df[\"x\"][line],\n",
    "                 df['y'][line],\n",
    "                 '  ' + df[\"words\"][line].title(),\n",
    "                 horizontalalignment='left',\n",
    "                 verticalalignment='bottom', size='medium',\n",
    "                 color=df['color'][line],\n",
    "                 weight='normal'\n",
    "                ).set_size(15)\n",
    "\n",
    "    \n",
    "    plt.xlim(Y[:, 0].min()-50, Y[:, 0].max()+50)\n",
    "    plt.ylim(Y[:, 1].min()-50, Y[:, 1].max()+50)\n",
    "            \n",
    "    plt.title('t-SNE visualization for {}'.format(word.title()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 10 palavras mais similares VS 8 palavras aleatórias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x648 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tsnescatterplot(w2v_model, 'jon', ['fight', 'battle', 'sun', 'snow', 'ghost', 'king', 'dragon', 'daenerys']) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 10 palavras mais similares VS 10 palavras menos similares"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x648 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tsnescatterplot(w2v_model, 'daenerys', [i[0] for i in w2v_model.wv.most_similar(negative=[\"daenerys\"])])"
   ]
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}