analise.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Classificador de poemas do Fernando Pessoa\n",
    "## Grupo Turing - NLP"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Passo 0: importando as bibliotecas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 71
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1192,
     "status": "ok",
     "timestamp": 1573684847920,
     "user": {
      "displayName": "Fernando Matsumoto",
      "photoUrl": "",
      "userId": "02497909482286168309"
     },
     "user_tz": 120
    },
    "id": "huhOdj9WJL-k",
    "outputId": "eed756fd-c641-4aae-da86-6bb723067c82"
   },
   "outputs": [],
   "source": [
    "import re\n",
    "import unicodedata\n",
    "\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.svm import SVC\n",
    "\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.stem.snowball import PortugueseStemmer\n",
    "from langdetect import detect, DetectorFactory\n",
    "import spacy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package stopwords to\n",
      "[nltk_data]     C:\\Users\\lucas\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "nltk.download('stopwords')\n",
    "tqdm.pandas()\n",
    "DetectorFactory.seed = 42"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Passo 1: carregar o dataset e filtrar por autores mais frequentes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 450
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 116748,
     "status": "error",
     "timestamp": 1573684968470,
     "user": {
      "displayName": "Fernando Matsumoto",
      "photoUrl": "",
      "userId": "02497909482286168309"
     },
     "user_tz": 120
    },
    "id": "inuQshNZJL-s",
    "outputId": "14c5baad-7707-4cbf-e6bd-9a6a4ca2669d"
   },
   "outputs": [],
   "source": [
    "dados = pd.read_csv('fernando_pessoa.csv')\n",
    "dados.drop(['id', 'titulo', 'data', 'bibliografia'], inplace=True, axis=1)\n",
    "dados.reset_index(drop=True, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nesse passo, optamos por trabalhar com os autores com mais de 80 textos, pois menos que isso poderia ocasionar um sobreajuste. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MrqnazFXJL-x"
   },
   "outputs": [],
   "source": [
    "dados = dados.groupby('autor').filter(lambda x : (x['autor'].count()>=80).any())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Além disso, percebemos que no corpus há muitos textos em inglês, para isso utilizamos a função **detect** da biblioteca *langdetect*. Isso eliminou certa de 625 textos."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████| 4383/4383 [01:37<00:00, 44.91it/s]\n"
     ]
    }
   ],
   "source": [
    "dados['port'] = dados['texto'].progress_apply(lambda x: detect(x) != 'en')\n",
    "dados = dados[dados['port'] == True]\n",
    "dados.reset_index(inplace = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Passo 2: Preprocessamento: stemização, lematização, etc. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ao analisar os textos, vimos que muitas palavras começavam com um travessão ou *underscore* indicador de diálogo. Para resolver tal problema, criamos a função `remover_underscore`, que lidou bem com o problema. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def remover_underscore(palavra):\n",
    "    if palavra[0] == '_' or palavra[-1] == '_':\n",
    "        return palavra.replace('_', '')\n",
    "    else:\n",
    "        return palavra"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Também percebemos que palavras acentuados prejudicam o modelo de stemização e de lematização, portanto optamos por remover todos os acentos. A função `remover_acento` talvez poderia ser mais simples, entretanto foi a solução encontrada. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def remover_acento(texto):\n",
    "    try:\n",
    "        texto = unicode(texto, 'utf-8')\n",
    "    except NameError:\n",
    "        pass\n",
    "    texto = unicodedata.normalize('NFD', texto)\\\n",
    "           .encode('ascii', 'ignore')\\\n",
    "           .decode(\"utf-8\")\n",
    "    return str(texto)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Por último, tanto na stemização, quanto na lematização, foi importante remover _stopwords_ do texto. Para isso, usaremos a lista de stopwords em português que o NLTK fornece. Além disso, também usamos expressões regulares (*regex*) para eliminar quaisquer caracteres que pudessem poluir o texto."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "stopwords_pt = set(stopwords.words('portuguese'))\n",
    "\n",
    "def remover_stopwords(texto, lista=False):\n",
    "    palavras = re.findall(r'\\b[A-zÀ-úü]+\\b', texto.lower())\n",
    "    sem_stopwords = [p for p in palavras if p not in stopwords_pt]\n",
    "    if lista:\n",
    "        return sem_stopwords\n",
    "    else:\n",
    "        return ' '.join(sem_stopwords)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 2.1: stemização utilizando SnowBall"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Aqui criamos as duas funções mais importantes do preprocessamento do texto. A primeira, `stemizar`, recebe o texto não processado e o devolve com as palavras stemizadas e sem stopwords. Para isso, utilizamos o algoritmo *Snowball* criado por Martin Porter. A biblioteca do NLTK conta com um stemizador chamado `PortugueseStemmer` e é ele que utilizaremos."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "r4CPtoN-JL-1"
   },
   "outputs": [],
   "source": [
    "def stemizar(texto_original):\n",
    "    palavras_importantes = remover_stopwords(texto_original, lista=True)\n",
    "    \n",
    "    stemizador = PortugueseStemmer()\n",
    "    palavras = []\n",
    "\n",
    "    for p in palavras_importantes:\n",
    "        novo_p = remover_underscore(remover_acento(p))\n",
    "        palavras.append(stemizador.stem(novo_p))\n",
    "    return ' '.join(palavras)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 2.2: lematização utilizando spaCy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A outra função criada foi a `lematizar`. Similar à função anterior, essa função recebe o texto não processado e o devolve com os verbos lematizados. Para isso, utilizamos os modelos da biblioteca **spaCy**, a qual conta também com ferramentas de marcação de partes do discurso, utilizadas na função para separar os verbos a serem lematizados. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "nlp = spacy.load('pt_core_news_sm')\n",
    "\n",
    "def lematizar(texto_original):\n",
    "    palavras_importantes = remover_stopwords(texto_original, lista=True)\n",
    "    \n",
    "    texto_importante = ' '.join(palavras_importantes)\n",
    "    texto_importante = remover_underscore(remover_acento(texto_importante))\n",
    "    \n",
    "    doc = nlp(texto_importante)\n",
    "    doc_lematizado = [token.lemma_ if token.pos_ == 'VERB' else str(token) for token in doc]\n",
    "    return ' '.join(doc_lematizado)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finalmente, aplicamos as duas funções ao corpus, gerando as colunas *texto_stemizado*, *texto_lematizado*, *texto_sem_stopwords* e *texto_l_s* (combinação de lematização e stemização)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████████████████████████████████████| 3759/3759 [00:36<00:00, 102.98it/s]\n",
      "100%|██████████████████████████████████████████████████████████████████████████████| 3759/3759 [03:40<00:00, 17.08it/s]\n",
      "100%|██████████████████████████████████████████████████████████████████████████████| 3759/3759 [04:08<00:00, 15.13it/s]\n",
      "100%|████████████████████████████████████████████████████████████████████████████| 3759/3759 [00:00<00:00, 4769.11it/s]\n"
     ]
    }
   ],
   "source": [
    "dados['texto_stemizado'] = dados.texto.progress_apply(stemizar)\n",
    "dados['texto_lematizado'] = dados.texto.progress_apply(lematizar)\n",
    "dados['texto_l_s'] = dados.texto.progress_apply(lambda x: stemizar(lematizar(x)))\n",
    "dados['texto_sem_stopwords'] = dados.texto.progress_apply(remover_stopwords)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "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>index</th>\n",
       "      <th>autor</th>\n",
       "      <th>tipo</th>\n",
       "      <th>texto</th>\n",
       "      <th>port</th>\n",
       "      <th>texto_stemizado</th>\n",
       "      <th>texto_lematizado</th>\n",
       "      <th>texto_l_s</th>\n",
       "      <th>texto_sem_stopwords</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Diana através dos ramos\\nEspreita a vinda de E...</td>\n",
       "      <td>True</td>\n",
       "      <td>dian atrav ram espreit vind endymion endymion ...</td>\n",
       "      <td>diana atraves ramos espreitar vir endymion end...</td>\n",
       "      <td>dian atrav ram espreit vir endymion endymion n...</td>\n",
       "      <td>diana através ramos espreita vinda endymion en...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>Fernando Pessoa</td>\n",
       "      <td>prosa</td>\n",
       "      <td>A REFORMA DO CALENDÁRIO E AS SUAS CONSEQUÊNCIA...</td>\n",
       "      <td>True</td>\n",
       "      <td>reform calendari consequenc comerc comissa esp...</td>\n",
       "      <td>reforma calendario consequencias comerciais co...</td>\n",
       "      <td>reform calendari consequenc comerc comissa esp...</td>\n",
       "      <td>reforma calendário consequências comerciais co...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Aqui, sem outro Apolo do que Apolo,\\nSem um su...</td>\n",
       "      <td>True</td>\n",
       "      <td>aqu outr apol apol suspir abandon crist febr b...</td>\n",
       "      <td>aqui outro apolo apolo suspiro abandonemos cri...</td>\n",
       "      <td>aqu outr apol apol suspir abandon crist febr b...</td>\n",
       "      <td>aqui outro apolo apolo suspiro abandonemos cri...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>5</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Em Ceres anoitece.\\nNos píncaros ainda\\nFaz lu...</td>\n",
       "      <td>True</td>\n",
       "      <td>cer anoitec pincar aind faz luz sint tao grand...</td>\n",
       "      <td>ceres anoitecer pincaros ainda fazer luz sinto...</td>\n",
       "      <td>cer anoitec pincar aind faz luz sint tao grand...</td>\n",
       "      <td>ceres anoitece píncaros ainda faz luz sinto tã...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>8</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Não a ti, mas aos teus, odeio, Cristo.\\nTu não...</td>\n",
       "      <td>True</td>\n",
       "      <td>ti odei crist es deus etern panteon presid vid...</td>\n",
       "      <td>ti odeio cristo es deus eterno panteon preside...</td>\n",
       "      <td>ti odei crist es deus etern panteon presid vid...</td>\n",
       "      <td>ti odeio cristo és deus eterno pantéon preside...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   index            autor    tipo  \\\n",
       "0      0     Ricardo Reis  poesia   \n",
       "1      1  Fernando Pessoa   prosa   \n",
       "2      3     Ricardo Reis  poesia   \n",
       "3      5     Ricardo Reis  poesia   \n",
       "4      8     Ricardo Reis  poesia   \n",
       "\n",
       "                                               texto  port  \\\n",
       "0  Diana através dos ramos\\nEspreita a vinda de E...  True   \n",
       "1  A REFORMA DO CALENDÁRIO E AS SUAS CONSEQUÊNCIA...  True   \n",
       "2  Aqui, sem outro Apolo do que Apolo,\\nSem um su...  True   \n",
       "3  Em Ceres anoitece.\\nNos píncaros ainda\\nFaz lu...  True   \n",
       "4  Não a ti, mas aos teus, odeio, Cristo.\\nTu não...  True   \n",
       "\n",
       "                                     texto_stemizado  \\\n",
       "0  dian atrav ram espreit vind endymion endymion ...   \n",
       "1  reform calendari consequenc comerc comissa esp...   \n",
       "2  aqu outr apol apol suspir abandon crist febr b...   \n",
       "3  cer anoitec pincar aind faz luz sint tao grand...   \n",
       "4  ti odei crist es deus etern panteon presid vid...   \n",
       "\n",
       "                                    texto_lematizado  \\\n",
       "0  diana atraves ramos espreitar vir endymion end...   \n",
       "1  reforma calendario consequencias comerciais co...   \n",
       "2  aqui outro apolo apolo suspiro abandonemos cri...   \n",
       "3  ceres anoitecer pincaros ainda fazer luz sinto...   \n",
       "4  ti odeio cristo es deus eterno panteon preside...   \n",
       "\n",
       "                                           texto_l_s  \\\n",
       "0  dian atrav ram espreit vir endymion endymion n...   \n",
       "1  reform calendari consequenc comerc comissa esp...   \n",
       "2  aqu outr apol apol suspir abandon crist febr b...   \n",
       "3  cer anoitec pincar aind faz luz sint tao grand...   \n",
       "4  ti odei crist es deus etern panteon presid vid...   \n",
       "\n",
       "                                 texto_sem_stopwords  \n",
       "0  diana através ramos espreita vinda endymion en...  \n",
       "1  reforma calendário consequências comerciais co...  \n",
       "2  aqui outro apolo apolo suspiro abandonemos cri...  \n",
       "3  ceres anoitece píncaros ainda faz luz sinto tã...  \n",
       "4  ti odeio cristo és deus eterno pantéon preside...  "
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dados.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Passo 3: criação do modelo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para treinamento do modelo, optamos por utilizar os 4 heterônimos com mais textos (Bernardo Soares, Álvaro de Campos, Ricardo Reis e Alberto Caeiro). Os textos classificados como Fernando Pessoa serão utilizados para treinar o modelo de *tf-idf*, mas não serão utilizados para treinar os modelos de classificaçã."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Fernando Pessoa     2327\n",
       "Bernardo Soares      503\n",
       "Álvaro de Campos     393\n",
       "Ricardo Reis         307\n",
       "Alberto Caeiro       127\n",
       "António Mora          95\n",
       "Alexander Search       7\n",
       "Name: autor, dtype: int64"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dados.autor.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Criamos então o DataFrame `autores_principais`, o qual contém os quatro heterônimos mais frequentes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Bernardo Soares     503\n",
       "Álvaro de Campos    393\n",
       "Ricardo Reis        307\n",
       "Alberto Caeiro      127\n",
       "Name: autor, dtype: int64"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "autores_principais = ['Bernardo Soares', 'Álvaro de Campos', 'Ricardo Reis', 'Alberto Caeiro']\n",
    "dados_principais = dados[dados['autor'].isin(autores_principais)].copy()\n",
    "dados_principais.autor.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 3.1: codificação dos autores"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para podermos montar um modelo classificatório, precisamos lidar com a coluna categórica *author* do DataFrame. O `LabelEncoder` do **scikit-learn** é capaz de lidar com o problema. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "le = LabelEncoder()\n",
    "dados_principais['autor_codificado'] = le.fit_transform(dados_principais['autor'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "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>index</th>\n",
       "      <th>autor</th>\n",
       "      <th>tipo</th>\n",
       "      <th>texto</th>\n",
       "      <th>port</th>\n",
       "      <th>texto_stemizado</th>\n",
       "      <th>texto_lematizado</th>\n",
       "      <th>texto_l_s</th>\n",
       "      <th>texto_sem_stopwords</th>\n",
       "      <th>autor_codificado</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Diana através dos ramos\\nEspreita a vinda de E...</td>\n",
       "      <td>True</td>\n",
       "      <td>dian atrav ram espreit vind endymion endymion ...</td>\n",
       "      <td>diana atraves ramos espreitar vir endymion end...</td>\n",
       "      <td>dian atrav ram espreit vir endymion endymion n...</td>\n",
       "      <td>diana através ramos espreita vinda endymion en...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Aqui, sem outro Apolo do que Apolo,\\nSem um su...</td>\n",
       "      <td>True</td>\n",
       "      <td>aqu outr apol apol suspir abandon crist febr b...</td>\n",
       "      <td>aqui outro apolo apolo suspiro abandonemos cri...</td>\n",
       "      <td>aqu outr apol apol suspir abandon crist febr b...</td>\n",
       "      <td>aqui outro apolo apolo suspiro abandonemos cri...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>5</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Em Ceres anoitece.\\nNos píncaros ainda\\nFaz lu...</td>\n",
       "      <td>True</td>\n",
       "      <td>cer anoitec pincar aind faz luz sint tao grand...</td>\n",
       "      <td>ceres anoitecer pincaros ainda fazer luz sinto...</td>\n",
       "      <td>cer anoitec pincar aind faz luz sint tao grand...</td>\n",
       "      <td>ceres anoitece píncaros ainda faz luz sinto tã...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>8</td>\n",
       "      <td>Ricardo Reis</td>\n",
       "      <td>poesia</td>\n",
       "      <td>Não a ti, mas aos teus, odeio, Cristo.\\nTu não...</td>\n",
       "      <td>True</td>\n",
       "      <td>ti odei crist es deus etern panteon presid vid...</td>\n",
       "      <td>ti odeio cristo es deus eterno panteon preside...</td>\n",
       "      <td>ti odei crist es deus etern panteon presid vid...</td>\n",
       "      <td>ti odeio cristo és deus eterno pantéon preside...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>10</td>\n",
       "      <td>Bernardo Soares</td>\n",
       "      <td>prosa</td>\n",
       "      <td>— A única vantagem de estudar é gozar o quanto...</td>\n",
       "      <td>True</td>\n",
       "      <td>unic vantag estud goz quant outr diss arte iso...</td>\n",
       "      <td>unica vantagem estudar gozar quanto outros diz...</td>\n",
       "      <td>unic vantag estud goz quant outr diz arte isol...</td>\n",
       "      <td>única vantagem estudar gozar quanto outros dis...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   index            autor    tipo  \\\n",
       "0      0     Ricardo Reis  poesia   \n",
       "2      3     Ricardo Reis  poesia   \n",
       "3      5     Ricardo Reis  poesia   \n",
       "4      8     Ricardo Reis  poesia   \n",
       "6     10  Bernardo Soares   prosa   \n",
       "\n",
       "                                               texto  port  \\\n",
       "0  Diana através dos ramos\\nEspreita a vinda de E...  True   \n",
       "2  Aqui, sem outro Apolo do que Apolo,\\nSem um su...  True   \n",
       "3  Em Ceres anoitece.\\nNos píncaros ainda\\nFaz lu...  True   \n",
       "4  Não a ti, mas aos teus, odeio, Cristo.\\nTu não...  True   \n",
       "6  — A única vantagem de estudar é gozar o quanto...  True   \n",
       "\n",
       "                                     texto_stemizado  \\\n",
       "0  dian atrav ram espreit vind endymion endymion ...   \n",
       "2  aqu outr apol apol suspir abandon crist febr b...   \n",
       "3  cer anoitec pincar aind faz luz sint tao grand...   \n",
       "4  ti odei crist es deus etern panteon presid vid...   \n",
       "6  unic vantag estud goz quant outr diss arte iso...   \n",
       "\n",
       "                                    texto_lematizado  \\\n",
       "0  diana atraves ramos espreitar vir endymion end...   \n",
       "2  aqui outro apolo apolo suspiro abandonemos cri...   \n",
       "3  ceres anoitecer pincaros ainda fazer luz sinto...   \n",
       "4  ti odeio cristo es deus eterno panteon preside...   \n",
       "6  unica vantagem estudar gozar quanto outros diz...   \n",
       "\n",
       "                                           texto_l_s  \\\n",
       "0  dian atrav ram espreit vir endymion endymion n...   \n",
       "2  aqu outr apol apol suspir abandon crist febr b...   \n",
       "3  cer anoitec pincar aind faz luz sint tao grand...   \n",
       "4  ti odei crist es deus etern panteon presid vid...   \n",
       "6  unic vantag estud goz quant outr diz arte isol...   \n",
       "\n",
       "                                 texto_sem_stopwords  autor_codificado  \n",
       "0  diana através ramos espreita vinda endymion en...                 2  \n",
       "2  aqui outro apolo apolo suspiro abandonemos cri...                 2  \n",
       "3  ceres anoitece píncaros ainda faz luz sinto tã...                 2  \n",
       "4  ti odeio cristo és deus eterno pantéon preside...                 2  \n",
       "6  única vantagem estudar gozar quanto outros dis...                 1  "
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dados_principais.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 3.2: divisão do DataFrame em treino e teste"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para essa tarefa, utilizamos `train_test_split`, com o DataFrame de teste contendo 10% do conteúdo do DataFrame original."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_X, test_X, train_Y, test_Y = train_test_split(\n",
    "    dados_principais[['texto', 'texto_stemizado', 'texto_sem_stopwords', 'texto_l_s', 'texto_lematizado']],\n",
    "    dados_principais['autor_codificado'],\n",
    "    test_size = 0.1,\n",
    "    random_state = 42,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 3.3: criação do tf-idf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para criação do vetor do *tf-idf* utilizamos o texto lematizado e depois stemizados de **todos** autores. Depois aplicamos a transformação aprendida em todo o corpus nos DataFrames de teste e treino."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "tfidf_vector = TfidfVectorizer()\n",
    "tfidf_vector.fit(dados_principais['texto_l_s'])\n",
    "\n",
    "train_X_tfidf = tfidf_vector.transform(train_X['texto_l_s'])\n",
    "test_X_tfidf = tfidf_vector.transform(test_X['texto_l_s'])\n",
    "\n",
    "dados_principais_tfidf = tfidf_vector.transform(dados_principais['texto_l_s'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 3.4: aplicando modelos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nesse notebook, optamos por utilizar dois modelos clássicos para classificação de texto utilizando *tf-idf*: *Gaussian Naive Bayes* e *Support Vector Machine*. Uma explicação detalhada dos modelos pode ser encontrada na nossa página do Medium: [Turing Talks](https://medium.com/turing-talks)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.4.1 GaussianNaiveBayes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia do Naive Bayes: 60.902%\n"
     ]
    }
   ],
   "source": [
    "nb = GaussianNB()\n",
    "nb.fit(train_X_tfidf.toarray(), train_Y)\n",
    "predictions_nb = nb.predict(test_X_tfidf.toarray())\n",
    "acuracia = accuracy_score(predictions_nb, test_Y)\n",
    "print(f\"Acurácia do Naive Bayes: {acuracia*100:.3f}%\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.4.2 Support Vector Machine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acurácia da SVM: 82.707%\n"
     ]
    }
   ],
   "source": [
    "svm = SVC(C=1.0, kernel='linear', degree=3, gamma='auto')\n",
    "svm.fit(train_X_tfidf, train_Y)\n",
    "predictions_svm = svm.predict(test_X_tfidf)\n",
    "acuracia = accuracy_score(predictions_svm, test_Y)\n",
    "print(f\"Acurácia da SVM: {acuracia*100:.3f}%\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A SVM obteve resultados muito mais satisfatórios. Além disso, durante os testes, a utilização da coluna *texto_l_s* aumentou em cerca de 8% a acurácia do modelo e por isso ela foi a escolhida para ser utilizada."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Passo 4: visualizando os vetores do tf-idf utilizando redução de dimensionalidade"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Afim de obtermos uma visualização do *tf-idf*, aplicamos a transformação individualmente para cada autor e então reduzimos o vetor a duas dimensões utilizando o algoritmo **PCA**. A utilização de algoritmos mais complexos também se mostrou viável, mas optamos por não colocar suas aplicações nesse notebook."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 4.1: aplicando *tf-idf* para cada heterônimo "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "bernardo = dados_principais[dados_principais.autor == 'Bernardo Soares']\n",
    "ricardo = dados_principais[dados_principais.autor == 'Ricardo Reis']\n",
    "campos = dados_principais[dados_principais.autor == 'Álvaro de Campos']\n",
    "alberto = dados_principais[dados_principais.autor == 'Alberto Caeiro']\n",
    "\n",
    "bernardo_tfidf = tfidf_vector.transform(bernardo.texto_stemizado)\n",
    "ricardo_tfidf = tfidf_vector.transform(ricardo.texto_stemizado)\n",
    "campos_tfidf = tfidf_vector.transform(campos.texto_stemizado)\n",
    "alberto_tfidf = tfidf_vector.transform(alberto.texto_stemizado)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Passo 4.2: aplicando PCA nos vetores do *tf-idf* e então utilizando matplotlib para visualizar os resultados"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nesse passo, utlizamos uma simples transformação para duas dimensões e então plotamos o PCA em um scatter do matplotlib."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAEICAYAAAC3Y/QeAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJztvXuUXVWV//ud9X6lHklOqlKVVEJCAnlYJnSJgAqhgR/YamKn40+6rzb+wMGQdOQyummbK+jIwMaLtr8r7QVi093cATYt2vlFDT7gJ2hEhWgKCGVCSFIJoVKpZx71SL0f6/7xPcu969Sp89x1nvMzxhnnnL332Xud13fNNddcc4oxBoqiKEp2kZPsBiiKoiiJR8VfURQlC1HxVxRFyUJU/BVFUbIQFX9FUZQsRMVfURQlC1HxV5QIEJFTInJjstuhKF6h4q8kHL+QDovIRRG5ICI/EZGlyW7XXCAiBSLyP0Wkzf9+3xGRbya7XYqi4q8ki48ZY8oALAbQBeD/jeUkIpLnZaO8Ph+A/wtAI4ArAcwDcD2ANzy+xly0W8lwVPyVpGKMGQGwG8Bau01ECkXkGyLSKiJdIvJtESn279vkt6L/QUQ6Afx/rm1/JyLdItIhIv/Ddb6PiMgbItIvIqdFZKdr33IRMSJyh4i0AviFf/unReRdETknIve72+xv3yMi0u6/PSIihbO8xfcB+IExpt2QU8aYp13nWiMi+0SkV0QOi8jmONt9lYi84j/fmyKyyfWaz4jISREZ8I9A/o/Ivykl01DxV5KKiJQA+CSA/a7NXwOwGsAGAJcCqAPwZdf+GgDzASwDcKdrW4X/2DsAPCYiVf59gwD+GkAlgI8AuEtEPh7QlOsArAFws4isBbALwKcB1AJYAGCJ69j7AVzlb997Qav+gVne4n4Afysi20XkPSIirveeD+A5AP8bwCIAnwfwjIhcFmO76wD8BMA/+j+fewH8LxHxiUgpgG8B+LAxZh6AawAcnKXNSjZgjNGb3hJ6A3AKwEUAvQAmALQDeI9/n4Cit9J1/NUA3vE/3gRgDECRa/8mAMMA8lzbugFcNcv1HwHwTf/j5QAMgBWu/V8G8Kzrean/mjf6n58A8Geu/TcDODXLtXIB/A2A3wIY9b/X2/z7PgSgE0CO6/jvAtgZY7v/AcB3Al7zAoDb/O+hF8BfAChO9m9Ab8m/qeWvJIuPG2MqARQC2AHgVyJSA8AHoATAa37XRS+A5/3bLT2G7iI354wxE67nQwDKAEBE3i8ivxSRHhHpA/A5AAsDXn/a9bjW/dwYMwjgXMD+d13P3/Vvm4ExZtIY85gx5gOgBf8QgCdFZI29jjFmKuBcdTG2exmAT9jPzf/ZfRDAYv97+KT/HB3+SfbLg7VZyQ5U/JWk4hfHPQAmQaE6C1rx64wxlf5bheHk8B9fFuVl/hPAXgBLjTEVAL4NjjCmNcX1uAPAH6OP/K6pBa797aDQWur920JijBk2xjwG4AI4x9EOYKmIuP+H9QDOxNju06DlX+m6lRpjHvZf/wVjzE3gJPvbAP41XJuVzEXFX0kqQrYAqAJwxG8F/yuAb4rIIv8xdSJycxyXmQfgvDFmRESuBPBXYY7fDeCjIvJBESkA8CCm/1e+C+ABvy99Iegm+o9Z3t89/gnpYhHJE5Hb/O15A8DvQBfXF0Qk3z85+zEAz8bY7v8A8DERuVlEckWkyH/tJSJSLSKb/b7/UdDtNhnmfEoGo+KvJIvnROQigH7QFXKbMeawf98/AGgBsF9E+gG8COCy4KeJiO0AHhSRAVCovx/qYH87/ga0vDtAS73Ndcg/AmgC0AzgDwBe928LxjCA/wn69s/6z/sXxpiTxpgxAJsBfNi/73EAf22MeTvGdp8GsAXAFwH0gCOBvwf/5zkA/g4cbZwHJ4q3hzqfktmIMVrMRVEUJdtQy19RFCULUfFXFEXJQlT8FUVRshAVf0VRlCwkZZNBLVy40CxfvjzZzVAURUkrXnvttbPGGF+441JW/JcvX46mpqZkN0NRFCWtEJF3wx+lbh9FUZSsRMVfURQlC1HxVxRFyUJU/BVFUbIQFX9FUZQsRMVfURQlC1HxVxRFyUJU/BVFUbIQFX9FUZQsJGVX+CqKksE0NwN79gCtrUB9PbB1K9DQkOxWZRVq+SuKkliam4FvfAO4cAFYsoT33/gGtysJQ8VfUZTEsmcPUFXFW06O83jPnmS3LKtQ8VcUJbG0tgIVFdO3VVRwu5IwVPwVRUks9fVAX9/0bX193K4kDBV/RVESy9at9PNfuABMTTmPt25NdsuyCk/EX0RuEZGjItIiIveFOG6biBgRafTiuoqipCENDcC999LP39bG+3vv1WifBBN3qKeI5AJ4DMBNANoAHBCRvcaYtwKOmwfgbgC/i/eaiqKkOQ0NKvZJxgvL/0oALcaYk8aYMQDPAtgS5LivAPg6gBEPrqkoiqLEgRfiXwfgtOt5m3/bHxGRjQCWGmN+HOpEInKniDSJSFNPT48HTVMURVGC4YX4S5Bt5o87RXIAfBPA34U7kTHmCWNMozGm0ecLW39YURRFiREvxL8NwFLX8yUA2l3P5wFYD2CfiJwCcBWAvTrpqyiKkjy8EP8DAFaJyCUiUgDgVgB77U5jTJ8xZqExZrkxZjmA/QA2G2OaPLi2oiiKEgNxi78xZgLADgAvADgC4PvGmMMi8qCIbI73/IqiKIr3eJLV0xjzUwA/Ddj25VmO3eTFNRVFUZTY0RW+iqIoWYiKv6IoShai4q8oipKFaCUvRclmtKJW1qKWv6JkK1pRK6tR8VeUbEUramU16vZRlFRlrl0yra20+N1oRa2sQS1/RUlFEuGS0YpaWY2Kv6KkIolwyWhFraxG3T6KkkpYV88zzwC1tcCaNUBNDfd57ZKxFbXcrqU77siOaB+NclLxV5SUwbp6qqoo/H19wKuvAldfzQ5gLlwy2VhRy/05u11qWVZKUt0+ipIquF09a9cCxgAiwJEj6pLxEo1yAqDiryipQ2srXTsAUF0NXHMNn7e3a5FzL3F/zpYsjHJSt4+ipAr19bTuq6r4vLoaKCgArr8e2LkzqU3LKAI/ZyAro5zU8leUVEGjbxKDfs4AVPwVJXWw0TdVVUBbm7p65gr9nAGo20dRUotsjL5JBvo5q+WvKIqSjaj4K4qiZCGeiL+I3CIiR0WkRUTuC7L/cyLyBxE5KCK/EZG1XlxXURRFiY24xV9EcgE8BuDDANYC+Msg4v6fxpj3GGM2APg6gP8n3usqiqIoseOF5X8lgBZjzEljzBiAZwFscR9gjOl3PS0FYDy4rqIoihIjXkT71AE47XreBuD9gQeJyN8A+FsABQD+NNiJROROAHcCQH2WLbhQlJQn05KhZdr7iRIvLH8Jsm2GZW+MecwYsxLAPwB4INiJjDFPGGMajTGNPp/Pg6YpiuIJmVbyMdPeTwx4Yfm3AVjqer4EQHuI458FsMuD6yqKkijr1Z0MDXDu9+xJT2s5095PDHhh+R8AsEpELhGRAgC3AtjrPkBEVrmefgTAcQ+uqygzaW5mHpzbb+d9JltyibReMy0ZWqa9nxiIW/yNMRMAdgB4AcARAN83xhwWkQdFZLP/sB0iclhEDoJ+/9viva6izCDbhvKJTE2caSUfM+39xIAncf7GmJ8aY1YbY1YaYx7yb/uyMWav//H/aYxZZ4zZYIy53hhz2IvrKso0si1PeyKt10xLhpZp7ycGdIWvkjlk21A+kdZrpiVDy7T3EwOa2E3JHLItT/vWrXRrAezk+vr4/u+4Y26ul2nJ0DLt/USJWv5K5pBtQ3m1XpU4EGNSc7FtY2OjaWpqSnYzlHQjyxfuKIqIvGaMaQx3nLp9lMwiy4fyihIp6vZRFEXJQlT8FUVRshB1+yiZj84DKMoM1PJXMptsW/WrKBGi4q9kNtm26ldRIkTFX8lssm3Vr6JEiIq/ktloAi9FCYpO+CqZTaJTIGQaOlmesajlr2Q2yUyBkO61BXSyPKNRy1/JfJKx6tcKZ1XVdOFMp9w7Xla70hFEyqGWv6LMBYmIMprrkYVXk+U6gkhJ1PJX0oNEWo5eXKu1lULnxssoo0SMLLxKka31clMStfyV1McK3fHjwIkTwPe/D3z608Du3XN3rXit1LmOMkrEyMKrFNkabpuSqPgrqc+ePcDkJHDoEDAyAvh8gAjwla947zrwSlTnurZAIgTVq8lyDbdNSTwRfxG5RUSOikiLiNwXZP/fishbItIsIi+JyDIvrqtkCa2tFJ+iIqC4mMJfUQGMj3u/UtcrUZ3rKKNECWpDA+cTnnyS97G0P9uK7KQJcfv8RSQXwGMAbgLQBuCAiOw1xrzlOuwNAI3GmCERuQvA1wF8Mt5rK1lCfT2wfz8tfosdAXjtOvCyFORcRhml0/oF2xG651HuuEP9/UnGiwnfKwG0GGNOAoCIPAtgC4A/ir8x5peu4/cD+JQH11Wyha1bgR/8gAJXUUHhHxkBVq703tJNF1FNN0HVIjsphxfiXwfgtOt5G4D3hzj+DgA/C7ZDRO4EcCcA1Ks/cDrZHCfd0AB86Uv08ff00OJfuRLIy/PedZBOoqqCqsSBF+IvQbYFLQwsIp8C0AjgumD7jTFPAHgCYA1fD9qWGWTCgqF42bYNWL06MR2giqqSBXgh/m0AlrqeLwHQHniQiNwI4H4A1xljRj24bvagcdJERZlk8yhQ8Qwvon0OAFglIpeISAGAWwHsdR8gIhsB/AuAzcaYbg+umV1onLRDuufLiRddLat4RNzib4yZALADwAsAjgD4vjHmsIg8KCKb/Yf9E4AyAP8lIgdFZO8sp1OCoXHSRIVPi9MonuFJegdjzE8B/DRg25ddj2/04jpZS7pEoMw1qeT+SpbrZa7TRihZg67wTQeSmZY4lUgV91cyRyA6ClQ8QhO7pTqBFuY992Sf6Fu8XIAVD8kcgegoUPGIzLf803mCUH3c00mVNAHJHIHoKFDxiMy2/JubgfvvB7q7gdFR4PBh4LXXgIceSo8/Syr5uEORKP93qizASvYIRENeFQ/IbPHftYuZIMfGmBUyN5cdwa5dvKU6XkzuzbUwJ3oBWioIn7peFK9I4pqNzHb7/OIXwMAAsz8OD/MP2t4O/Od/Ru8GSob7KN7JvUS4jbIx9FBdL4oXJNmtm9mW/9mzwNAQLX8AMIa38XEgPz9yKzVZ6RXitTAT4TbK1tDDuRyB6Are7CDJbt3MFf/mZmBigr5+Y5gD3vjTBRkDfOc7wNq1Tr6YUB/2bF/SP/4j3UhnzgB1dcCOHcxB4xXx+rgTIczJ9n9nAm6xLywETp9m4rpszeOULSTZcMpc8d+zhwJ0+DCFP5DhYeAPfwDeeQfo6Ah9rmBfUns78Pzz/JMuXgz09gJf+AL3ed0BxPqnT4QwJ9v/nYpWcjRtChxVvvAC0N/Px9aNBqTeJL8SP0k2nDLX59/aClxzDd07ubkz94twe18f8OqrwF13ze5rC+Z7f+UVoLQUqKzkn7SyEigvBx591Pv3EiuJCI1Mpv87FUNho21T4JzJ2Bgwbx7w9tvOMdngRstGkhy6nLniX1/Psn8rV7IDyMmZPgKwfzRjeNzrr8/+Jw32JQ0MALW1048rL6cLKFVIlDB7UeovFlJxsjnaNgWuGbCP3caGutEykyQHDmSu28e6IzZu5ARvezurPxlDi992BAUFtNrHxpw/aeCHH8z3vno15xTc9PfT959KpEJo5FyRipPN0bYpcOi/Zg2wbx8NiakpDSPNdJL4/8xc8XcL9uAg/4CnT9NiLy5mFBDAx2Vl3B/qTxr4Ja1f7/j4y8sp/P39wAMPBH99Kvqmg5Eu7QSS7jP1pE2BcyYFBcCll9KIaGtL3Upi6fQ7UYIixqRmwazGxkbT1NTk7Umbm1kK8MUXGQVUVgYsWsSRwDXX8I83NgZUV0f2o969mz7+cNE+7kk996RoqkVwpEs7LanY3ljalG5Cmoqfu1ek23cRBBF5zRjTGPa4rBJ/S3Mz8Pjj7AQWLAA2bKDf/+RJuoVWrvT2R71z50xr0D7fuTPed+Md6dJON6n4Z03FNsWL+z2dPEljZ9UqZ3+q/04iIUM6tUjFP3PdPqFoaAC+/e3pP+jFi/mDLiz0ftFFLL7pZAhIKvrQw5GKcxqp2KZ4CAxH3b8fOH+e7s7qah6T6r+TSEiXXFoekZ3ibwn8k95+O91AbioqgIMHadHEKsTR+oGjWVHsZSeRij70bCYVRhDNzcDdd3Mx46JFwOWX8763FzhyxBH/TPidpKPxEweZG+oZC8Hi+VtauBAsnljyaON5Iw0X9DrOPVVSJiupsYbBtqG7G1i4kAsjX30V8Pn4++juzqzfSZYVysluyz+QYKtVm5roCnr5ZW5bs4YhnnffDaxYEZlFFm2ahkgtEK+HqamSMjmcxRvvfi/b4vXrLKnggrBtWLSIwl9czO09PYx2a29P7YikaEn2avUE48mEr4jcAuCfAeQC+DdjzMMB+68F8AiABgC3GmN2hzvnnE74hsL9py0oAH7yE84HFBdznUBHB0NHJyY4UVxWxj/AJZfwuRdD80gnXm+/3UkDYJmaYnuefDK+NiSLcJNu8e73si1ev85NKny3tg09PVzRXlREQ+jsWaCxMe0mQiMiFVxtcZKwCV8RyQXwGICbALQBOCAie40xb7kOawXwGQD3xnu9Occ9D7BzJ/+8p05xXcDUFC2gnBx2BkeO8HltrTP0Deabj/YHtXUr8MUv8k83Oso/nM8HfPWr04/LRB99OIs33v1etsVNYDRMbW18bUiF79a2obqaodBHjji+/0wUfiDzJutD4IXP/0oALcaYk8aYMQDPAtjiPsAYc8oY0wxgyoPrJY6DB7korK+Pwm9TQ09NOXmBhob4hz90iK6ho0enF4qJ1XcbmIwuWHK6VPLRe1XvIFyJRLu/s5MrYX/0I35PBw9G9vpo3sszz/C8XV2hz2W/4+PHgRMngDfeAF56iUkFY21DKny37jb4fMB730uL/1vfyhqBzGS88PnXATjtet4G4P0enDf59PbS1VNRQeEfHnb2XbzIexEeI8IRgTG08A4cYOcwPg6sW8f5ASAyK3DPHh7/J3/ibLtwYeZrku2jt9buwYOcFF+3jqtT3SMg+34iHfXU11NEz5zh51dRMT2mvL4eOHaMwlpUxHDDvj5+V83N8VvMbpdNbS1f+8ortHyrq4Ofa88eVoo7dIhtqqykUfDyy1xHUlMTvdWe7O82VdqQbqSR28gL8Q9ikiKmiQQRuRPAnQBQnwquC5vzp6TEsfQtU1MUfHtfVgacO8fjzp9n5zBvHoXozBmK1fLlnDD2+fjjmO2HEk3IWbhh6lz9GN0ieeECP4PDhynGNTU85vHHKYLRFMFZvx54+mmep7ycot7aCvz5n3P/1q3Apz/N6xUVOfma1q/n+/SyAM6aNYxuEQHeeotzQMHO1dpKX3xREd2B9vsdH6erpLAwtonDVHBBpEIb0oVkFX2KES/cPm0AlrqeLwHQHsuJjDFPGGMajTGNPp/Pg6bFyYYNjggPDAB5edMn4NzFYXJyOBF85owzArCdw9QUJ8mGh2lFnjhBQZjNHeRVyNlchgu6RbK/n0JbVOSkIq6o4GKgaLNuHjoEXHUVO96uLoo/wDQazc38E11yCc/f30+xveYarspubXWs1bEx4LnngF/9ip13pLjdRjU1wNVX83l7e/Csi83N9PEfPEhX1MWLNASqq9l5zfa6YCSjVKjiHamYZTYEXlj+BwCsEpFLAJwBcCuAv/LgvMln61bgtdecimDFxRTQQHJyaPX7fBT9vDwKv/XXGsOO4dQpHj88TP9paytFqqKCi2fsD8WrkLO5DBd0j04qKvieioqcTquvj59BtP731la6jsrL+Z4rK9lRdnc7VtSGDcGjodyd4+AgcO21zucXqQUW6DaqqeH1r79+ZuoC27nW1tLCt/M/1dX8TVxzDbO/RpLyIM2sRiUIabZILG7L3xgzAWAHgBcAHAHwfWPMYRF5UEQ2A4CIvE9E2gB8AsC/iMjh2c+YQjQ0AA89BGzaREEHKEZFRXTv5OWxQ7AFYxYv5v3YGEV/fNypGwzQRZGXRzHav5/by8spnL/4BfCb33CScc8eYPPm+PN8B05+dnXRQn3mmfgtS/foZM0avre+Pke0L1ygBR/tCMae9+23HTfK6CgjTNydY6jJ0HgssGgmWu11Vq9mR5OTQ0NhYIDzH3l5kU/QemU16ugheaTZIjFPFnkZY34K4KcB277senwAdAelHw0NjN656y5aYk1NFOSJCd6mppyO4Z13aCm2tXEC0C38OTncZn3kdq5AhNvPn+cP5fLLuX/v3ugzQQLTtxUW8pxVVRT+V17h9WxoajyWpXt04vPR537oEK81OsoqZx0d/EzWr6dbJpIRjD2ve1XpyAiwbBk7rna/R3HzZl4v2ERkPBZYNJOc7uusW8fJ3SNH2MbVq6ObX/HCasyE0UMaTZjOIM0WiWVnVs9YaW4G/vIvHb+2FW6Aft61aymEv/kN/dH2sxWh33l8nB1FXh732QpjY2PcNj5O0Zic5OTixo1MQBesHYGLiIJlJD1xgtdesYLC2dfHY66+mp1UvJkYZ+uA3G1raeFEcDSL4ALzySxaxBBa60aybp/ZRC1R2Um9vI4X50rHrKxuMiGrZgp0XprVcy5oaKAbwkaZWOEH6GOur3dcQIWFFFpbKtKGiZaUUOgHBij8BQV8PDTE1737Lo+ZN48pp3fvnm7hbt0a3Jff3c3HjY3OtpUraYVXVdEara2li8ZG48TrjwwWCbJz5/S2rV7NDrGqymn7I4+E/mM0NDCW3ArBwYMUfmPY/nBzF4EWmLsD2rnTuz+kl5aeF+dKM5/zDFIhpUW8pFF0lIp/tPT1UZiNmR7fL8IVudaPD7BDKCigAE/517cND3ObXSpfUMAOpbfX6Szsa6uqWHxm06bpw/j+/pk/sNFRdiD79jnx8ZddxtGEtfpiiX+P1pKxAtTZyRGSnQfIyeHoJFKXhNv9Em3H5X6tXYNgXU9eukK8jIP34lxzsSo4kZZsundeaYaKf7TU1fEP4fPxjzU56fj9+/oowvX17AhOnuRzN6OjvBUV8d52GADPNTDghI5evMjj3nyTwmfT57a2Or58y8QE5w2qqpxJ5JdfBq67jvtjsSwj9SEHpjbo6KBguxdhdXREn/LAbUVFK2r2tTt3cr5grqxJLy29eM/ltc850XMIqZDSIovQlM7RsmMHxXloiBb7+DiF1+fj/blzrBV8/jwnPfPzg59nZMRxCbmxz6emnE5ieJjpAn78Y8atd3bSn++OSBkb44gkEJsWwlqW0UQQRRKBEriWoK6OkUw29NMuwrIJ8NxEatXNFoGzfn34yJZY0j2EipjZvZsjsVWreL87bI7CxBHLdxyKRMetp0JKiyxCLf9o2baNfvl/+idn9WppKYU6J4fhni0tFL+pqemLwgKxrqDZsBPKNhpocNCpNibCa9qUumvXcmRw9Kjj9tmwgccEDt3vuScyQWhtZecV6EpyC2egn3bVKj4eG3MWf11xBVfI9vRMP380bqf+fl63spLv60MfYkRUOKs03kI6x45xRfEll/CzePVVRiEtXkxX3Re+wNcFq92cDLwciSTaDaPpJBKKin+kBAroo49On4jt7GSnsH8//fiTk7R6wwl8KEpLeZ6ODgr+4CCvYYWsrs5J72sjPTZtcl5/4QI7n1iH7gUFHGnYVAuBriQguEAsXUqh37LF2Wbz71y4EJvbqaHBec1sk97ATHdOtK4Q93k7OzlRbBfstbSww6+p4edaWcnXPPro3It/czNDju36kKuuArZvn1thTIYbJo0mTNMdFf9ICOb7tHH4AAXjpZcoeLm5dL/Ylb2BPv9IEaHgLl5M6354mB3B0qXA/Pl8/uKLTsqDYCJ34gQFbGjIKcFnJ0wj8XkHyyQauD2YQCxZMlPo8/KAL31p9tj8YIQS+GCdzsgI8MMfcp872qqkhPeRFB5xn9cuNCsq4shjZISuvrNn6cYC+B2dOTP7e/CC3buZ4ruriwbB/PnslNvamOZ7rsQyzeLWlehQ8Y+E2URo1y4nKqe2lsJh3TFW9K3fO1rsArE/+RN2KCdOULgWLOB+ET52i3hpKS1zY+h+EaHwu0vwXX01O4JIhu6jo1y5GsyVZAkmELm5swt9NBZyKLdDYKfT2cn3Xl7uuKoAtt8mVos2vYONVLKZXefNcxadWfr7OQKbK5qbGfHV1+d0OD09/A57euY2DFLdMBmNin8kzCZCzz1Hcamqos/99793wjrdSd0WL+bk5C9/yUnh/Hzeh1tgl59Pa2/jRkbPFBc7IaYXLlCMnnmGwnfmDBdzfexjFIp9+7jqNLAE39tvUwwjGbpbIQx0JS1e7DwPJRCBQh9t2GAot0Ngp/PGG3x8xRVcZWvDbY8eddofiVC6z2sjlYzhNbu7+T3k51P0p6Z4/8ADoc8ZD3v2MKgAcHJGAbxubu7ch0GqGyZj0WifSJgtZ4cxTiRJdTVdKnaSNi+Pf05jnDQLJSU8fvVqbguFz8fzbdjAVb5XX02Bf+MNTp6eOcNhf3k5t7W00LVhozLGx3mMzbszPOwkSIs0giLS6AsbUvnkk7wPJhaxZBgNdf3AyJaxMXbENue+ddfY7y3a9A52pGcMLfujRzkHUlPDEVZLC7/jr389+GjGy+I2Ph+vNTHBbXl5HNFF2okrShDU8o+E2XyfNnFZRwet/o4OR/SLi53FXOfO0QKvraWV3tnpLAwLZv2L0FUzOurk/R8a4vkKCvh4cpL7BwZ4fH4+8LOfOUvji4vpFqiuZmfz298yYigvjyOCSPBy2B/L6k17/V27OMqyE53u/fa1dsIbcLKM2sdAdBOV7vPaVBMTE/wOr76an6kdkcwm/Pffz462txd44QXg3/8d+MhHop+kra93akefP8/PwOaN8vk0DFKJGRX/SJhNBAHg85+n5W3Ffnyc4v+nf0qR/dnPKBof/jBLDq5YwXUA9g8cDGNoaeblsaOwi8rsKt6xMWdlcFUVzzc2xk5g2TIK3/nz7BCOHeMiseFaWRbiAAAgAElEQVRhWqxVVXQ/7d8fmRh5NeyPJ2wwkvTM7g76ssvo/wecPEDRTlS6XVR9fQwtdbu7QrV91y5nZNDb68wD/fa3/P6CLZJ7/HF+JyLs4O66a/pEfmMjfxOtreyIbryR8yrqklFiRMU/UmYTwcFBp1jIvHm00nJzWcaxtpZW/w03cH9FBZ/bY3NzKQbBsDWD8/Lowz52jKOFhQspKBMT3G9XCk9NTS9aUlTEeYb2dgqfdTnZamNTU+y0EpX1MdawwUhHDIEd9KZNTrTP4sUU70jyCgEzo7sOH2ZnsmmTs8o6VNv37+f3293N7y8/n+3o6XEWSblHFl/8Iif07SK9ffvoynrooenvq6iIczrplOlSSVlU/OOlr49uFbuY6+JF/ultqoWbbqKVbqs8tbQ42SkrKynEAwOzn7+720khkZvr1AgYHXVWFOfkMPzPVreyUTnj4xSU3l5ue/ddJ6vo6CgFKVCM5opYwwa9KGkZbZqCwA5n40aGVr7+OnDzzeHb7q7fYOd27ERtYNv37GGnUF7uTMqL8Hu334tOuipzgIp/vNTVUVztgp+yMoryihXORJ/bsisspPAODjJUM9w6ABvpAbATGB6m2NvO5tgxWvVFRVyF2tNDcTp4kJEv1dW0XEdG2PmMjzuuqdzc6WIUKhonkvoBoSzShgbm4H/0UU5E19UxVUa0lbWA6BcaRTvfENjh1NTQ7fS730W2VuCqq9hZ2JGdu6NtaaGx4L7W6Oj0FBR2oloTmilziEb7xMuOHbS2e3tpnff28vmOHdzf0MCFWeXlFN3KSpYpXLaMwmALukSKMbyOLSZjk8G9/TbnFE6fdkIRT5+m66ewkP7i3l52Ona+4eRJhoqePMmFRLNF4wRG6hw/zonOj3yEcxr5+dx3//3A5z4XPMKluZkL4977Xs6HDA8D991H33asET+REm1+n2DRXUVFwMc/HjqiybJ9OzOIlpTwPBMTTpru/fv5nbivVVg4fe2AHTFoJI8yh6j4x8u2bQz3q6xkREZl5czwv9FRugu2bOHEr635a1cBR1tQx1r9tiMAnLKRp05RzEtL2eG8+CKFzhaVn5ykFToywtv587TCv/IVp9JYYBIvt+Xc08PFWzbOHaCgdXTQqn3jjeChnPYco6M8HqCr6vXXZ4Z8usMkvShpGW15vXg7nIYGrrytq2PHv2gRJ+xraxktdOjQ9Gv5fPw8h4Z46+/nazSSR5lD1O3jBdu2hV656nZdVFezsPfu3U5G0EjIyXEWj7mLyADT8wdNTVHY3nyTt5IS4IMf5L558zhKmJpyQkVzcrh9fJziumqVcy63dWzdIEeO0AoG2Hl0dTk5h5Ysmb7WAHBcK9aV8vLLTm1eYyh07nmHUKk0YvV7Rzvf4EWIa0MDXX+2tq9lamr6iMN2FO5on02bnGgfRZkjVPwTQaD4FBTQCmxv50jB1gUIRWAt4FBMTHBSd2qKfv5XX+XrrfBbpqYo3Hv3MiImVNbNwJQHU1O0UouK2MF0dXFSe/ly5/XuzsN2gPb1gJM2wR4XWL4x2lxEsxGNmMeaATUYkc5XNDQEL9epKHOIJ24fEblFRI6KSIuI3Bdkf6GIfM+//3cistyL66YNwfKsf+lLTr7+ykqnCLwb91yALfCSm8vOIxy2uPzUFEXfXU0skPPneZuYCO7qcLtBbMoDu4p5YIBi7a5CZnELnT1HQQH9/TZHzpo1PK6gYGbh9ldfZYfiRRrhcKuQm5s5X7F16/R5jHCrkEOh+emVFCZuy19EcgE8BuAmAG0ADojIXmPMW67D7gBwwRhzqYjcCuBrAD4Z77XTimDheg8/zM4gL49+37ExZx2AjeixAg44awNsmGek6aLt5PBs2Iii976XHVMw69hazlVVnDguKnLWIgBOgfrz5x3Xk9u1YjvAxx/nPMSCBYyKKSjgcXYBWjy5iIDYyg5aV9PRo5yHAOiCueaa+EJhNTGaksJ44fa5EkCLMeYkAIjIswC2AHCL/xYAO/2PdwN4VETEmGhnOjOI5mb6hI8epfDV1DAE0oZgWveO9fOXlXHl6pEjFNm8vJmWdjDcOWFCHTN//vSwUnc73eL14IPcfsstzoK2ykpObtoRxmzhkA0NjIQBKK6//z07gHvv5eIrn48jgVde4THuXESRrM6Nteyg7dTGxjiysSOuI0f4mf/oR7HXsNUYfSVF8UL86wCcdj1vA/D+2Y4xxkyISB+ABQDOug8SkTsB3AkA9Zkc5rZ7N6NrxscpUufOMVImP58TruPjdHdY331ODi3rjRvZWQBO/iBbJnE23KMDO5JwY4Xu2DH6/Ds6KOCHD7NGQWUlOym3mG7ezOuXljp580+dYgeyYYNTYCYQtzhv3uyMDgDHP24nxI8ccXz/kU72xpI/CHAmo21OoOJijmw6OzmXUV6emBq2ipJAvPD5BwtSD1SjSI6BMeYJY0yjMabR5/N50LQUxOZnF6Glawu2fOxjwHvew4VaFy86cwA5OewUPvQhjg5yc6eHakYzeLJROhYRbsvJ4cRvZ+f09QBHjlDUA8M/H32UE9Y1NWzb1BQ7gYICiv9shKoJ6/aP+3x0QTU2At/6VuRCG0u9XsAJBb38cicD6vCws/J648bE1LBVlATihfi3AVjqer4EQPtsx4hIHoAKAOc9uHb6YfOzV1RQfK2V2dbG52+/zQ6hooIdgM3e2NLCAu4XLzrnimRxmLX0bfSPm8CJY5uCenyc1aomJmh9u6mooHtqwwYeX11N10hdHYUz1GRmKHH2ovh4tPH8FtvxFBY6WUPPn+ccy7XXOhFH7vYmA6/SRCsKvHH7HACwSkQuAXAGwK0A/irgmL0AbgPwKoBtAH6Rtf5+m5/dlgQEKP62OtNVVzEEtK+PbpV332VI5dKljNu3fv5oFoeJBJ/0HR2lgOfnO5FEk5N8DjhzDm76+ij0RUWOe8ZG69x008xsle75AlvXYLbQx3j946Hi+UNNBAdOzH74w06dYOuWCtbeRBLrfIaizELc4u/34e8A8AKAXABPGmMOi8iDAJqMMXsB/DuA74hIC2jx3xrvddMWm5/98GE+t3lc8vOd1A8298u+fRTls2dpgV+8SCGPJNbfTahOIjBpXEGB07EYQ7G5cIFtfuUVp4BMezvdMtde64jsXXc557ViNTHBkcL+/Xy8eDFHDXNREzZU6m3blqNHWRvgn/95elrk2TqeYJ3Jhz5EyzuWSeBYopGA2OczFGUWJFUN8MbGRtPU1JTsZnhPoCj29FD4bc1b96KgH/2I+2yq5t//3onIEQkenRMNtpiMjTACKOzW/VNcDLzvfcAf/sCJYBFa/aWlTrvXrqWYByZ6O3mSr29vn15Va2iIKa5HR2OLnomFnTs5od3URHdOQQE7vclJjrRCFUEPFOv167kozhbNsR1CJBa423qP9rW3306LP3C1cFvb7BPsSlYiIq8ZYxrDHacrfBNNYH726693BHD16pkrgfv7WcR9/35uO3t2ZiGY2SqChcJO9rpHEAUFjuunuJiuptpapyLVxAS3l5TQRQVQ+K3/2e2W2L+fUUw+n+Peqqig26q6mq9JFK2t7GgHB+l6ys/n5zU0FL4IeuCIYOfO0BZ4KMs+Huvdi+ymiuJCxT8ZBAqKnchrbaVVbWPlN26kaBUU0OKz5RsDxd4WC4kWW1fATvKWllKsCwspNEuXUmz6+7l/eJhtLSjgugP35GegsJWU0PofHKS7auFCdiw+H9NNx+o2iYX6enZGIyNOJlS7dsKWyoyUUPUFwvnl46lmFms9hHDE6oZS0h7N6plsAtMl2xq999zDfC8PPUSB6u6mwJeVUVjz8xnBU1AQeXI4d3SPLSxvs4GWlNAfbyOQbAF4gKI9OOikiBDhSODMGbb/9tvporJpibu62GHYJHQ226hdyfvOO9EVco+XrVv5GQ0O8t62a3iY99FYz6EiikKFsoZ7bTi8iIYKJPC3l4jvIlXJwkgqFf9kE04wGhooVrW13D82xlthIf+wNje8jbO3vns31k88OemsGygtdSx4YyiMHR0U8OPH+fjNNxn7D/AYEef8NpfPyIjTab38MoX/yBHH+rVtLizk7cwZtnm29zsXNDRw3UBREcXfGL5/W1s3mlw7ofL1hFtn4EWq6FD5iaIl3G8vW8jSTlDFP9mEE4zmZuDnP3fi+21Ipp1Avewyumcuv5yRQu4EcTk508sIFhRwwrasjMVGHn2U0TplZTzf+DitcivWg4PAs89yW36+Mx8wPOws7LJhohs38jqvv85RgTEcTfzZn7GQvc/Htl1yCa9t6eqiG+iZZ+bW4po3D/jkJ7mQrqyM72XFCk5YR5uuYTYLPJxlPxfWezzEuigu08jSTlB9/skm3ETenj1MgtbWRmt1cJBCPjJCcT18GPj854GvfY2uFje2WIx11YyMOGmbp6aYWO4DH6AwXrhAUbcuHjvpm59PMbdZPOfP5/auLm6zIwF3qUNbnezqq50FUu73aB93dTF81I5s3D5y+95j9UUH+rILCthZffSjzjHuNkXj+54tLDQSv3wq5frRSWQSz1xMGqOWf7IJ5wpobWVEzfAwxdcmHhsb4w/2kku4uKq42BFimw/Iruq11rmd2L3kEh5/7Bg7hJoaRu/YSeSJCVrHZWV8nXUR2ZKRQ0POCMSNLXX41FMckdg5hdnSQ7/1ljN5vWaNY3E9/nj4YXhzM9cVbNzIz+cTn+Dz229naub775/++jNnWEc52Ofs1bA/1Sz7cCQz5XQq+djjmYtJYzTOPxUIZXXu3Mk/5G9/6/jkx8Yoxpde6hRpf/rp6dW5zp515gHGx50wx5IS+tyHhylQS5bwHK+84qR0GBjgaMB2GEVFHFUMDzuLzOrred7JSbqdenr42g0b2GkMDNBdZFcE79jBamfNzU7VqtZWtl2E8fcAjx0cZNlLt0VqLVQrFPffz5QX8+bx+LY2jkr+23+jG6m/nxWxqqud19sw08DP2X7Gs11vrr7bVCAZ7YtnvUM2tCdOIo3zV/FPddyLwn77W7pKpqb4Iy0roz99cJDuH+vjn5igGFur31bgsr79mhp2Ij4f8NprjPIxhuceGOBramr4OCeH4t7RwfZceimFdcsW4MABLpwqKeH5Ll7k9TdupBiPjFCQz5/naKG+ngK8bh3P84MfsIPIzWXbAB43NsY5DPs+L7+cIxO7oGnnThZcATiCOXXKGRmtWuWsmC4pYQcAhF4QNVcLqDJMVDxjLjvbWEn1TjoKdJFXpuAugnL+vJNKuaCAAldRwZXBtsiKLfJi0zOsXEkRHBriqKCqioJZUcHJWWP4x7PCfd117AROnWJHUVZG4bedQmcnO4YLFyiOV1zBVbxjY7TCJydZgau6mq+5eNFJJX34MDuSw4fpvpqcdMJDAb4nW1+4rY2if+4c8MMfss3LlvFP2trK92InK0dG2PaxMQpsRQXfr3so39fHY9zrC9av52f3yitsa1kZO8I1a9iWgoL41iNoSobgpKKPPZXmYhKEin860NBAwV26lNa6zeZpXTeTkxS2RYtoldvUC+XlXFzV1+cUV7/5Zor1T37CjmLBAicENCeHYmirWZ0/T1dQfj4jY6amuN6gtpbiOzoKvPEGO6SREYqnnSA+d46dlBVjW/JxZIQ5jJqa2JG4M4lOTrLDsakrzp/n9Wy94NpaWtIlJXy/NjleUZFj+duRwq9+5dQa7uvjgjNjKOhLlnC+4+mnKfSTk07nmJfHzs/n47lsSG0sidRSUeRSAZ1oTglU/NOFUNlAV650CsLn5VHEFizg8W63R3MzO4CXXqIQzp9PwT93zrHOjXFq8gIU88lJXn/tWub6mTePYllR4bzWnSG0qMiJRgKcamJFRU7R96NH+T6Ghpz6xCIcKcyfzw7i9GnHqq+qYvoL67tftIg+fxuzf+ECX2cnmleuZGdpq4rV1TnnAfhZlZdzFFJdzffS0cH3s2wZ27VxY3xWu4pccOZqtbISFRrtky5YAbPFRoxxfNs7dlBg3/teFoVZtoxCfvnlzuv7+jgZW13tlIDs6qIIjo87i8eGhqZf11rL1i9fVkY3zG9+Q3Hu73dSOl+86Lh5JiZouRcVcVtvr2PBd3Y6k9BjY07KBbvaeHyc2/Lz2V5r2Xd1USxGR7nyedMmHpufz4neW27h86oqJmvbtcuZIxgbmx7T3tdH8e/vd9xbl15Kt8/NNzsjFjfRWu1awD046RYVlaGo5Z8uWGtp3bqZ2UC3baNVbCesrriCwmxDLd2W1a23OrH+buyKXXeBGBuvb5O/dXayvkBvL88nwvPbqmL5+c62pUuduQSbIbSggKOBjg6Ke06Ok+zNRinl5LDTuHjRWVRWUED31YEDTiWzPXsY2rlrV2SfX6AVXlHB91Fe7oymRkYcS7SuLnTtgUjQAu6zk4U+9lRDo33SiWgiEoIdCzBD6Gy5gOy6AOuGsa4gu8DLrheYmuLNnseK/tgYfegf/CAt9q4upj8eHKQbx0YanT3L48+c4bb8fIp9fz9dOENDFHzbGdn6uidO0L117bV83WyRM7O9d3fkTUsLw03XrGFHad/X+vV8r5s3x566WVGSiEb7ZCLRWEvBjg0XRldcTAEcG3OsfVvJKz/fqR/gTiltRX/hQor2/PlObH11NecdfD4u/rL88Ie8X7HCKRg/MUGr3qajKC3lvc0makNNS0udJfjATB98qMyabit89Wp2CocOOZPUlZXOdptiO1lWewaFHiqpiYp/phFKNGzK6P7+mfn/3auCbShnXx/F3dYUbmubWdbRhpQODXHfW29xYth2AFbM3dhtGzcyLLSoiFb9+DjPVVvLidepKUYITU2xbcuXs4N45RWWkPT5Zvrg3eGVNslcdzdw990sBh/YAW7bNvtnmSzXhJZsVBKATvhmEuHSFNTXc87A+vLdLFjAieInngA+9Sku4rr0Ulrnl10W/DVurNVvDMtPdnQ42UGPHAEee4zlE48dY6SOrRtgC6ZPTXHStbqa+xcscFIwFxayA7KFbIqKnNrBgT54m6zM5g0aHuaopLs7upQNyUw/kKWJxpTEouKfSYQTja1bGQK5fv30vDw2Bj8/n26QrVsZJWPTNLz7LoU3J8TPJS+Plvv111Oo9+0Dfv1rjjJycijChw7R5fP733PyuKODr/nwh9npXHklxX5oiG6YwkKOVFavdkS+o4Pbu7uDR87YPC1HjrCTKC7mcUNDXFtw993hhdzrFL/RdiSabVNJAHGJv4jMF5Gfi8hx/33VLMc9LyK9IvLjeK6nhCGcaNjokz//c2dS1udz0jWfPs2snN/4BrB7N63n4WFOxg4PT68ZbOPyAd7b1cbr1jGk1BadHx93In1sSKfNq9/UxJTPO3fS/fLVr3KFcX8/BTcnh9FBtojNsmV8/dmzHB0Ec4PY8MrubnYS587RXVVWFvkIwEvLO5aOJEsTjSmJJV7L/z4ALxljVgF4yf88GP8E4NNxXksJRySiYROZXXcdC6lbn76dQH39dQrzww9zInZ42Mnf7ybwuZ2w7exkUZfhYaeDsGkebBjo6CgnV8vLWVPA3bbt2zkCuPRSJ77/+HFGBuXmsgNobKT/Ppj/23ZwixaxkxgYYNjmwoW87qJF4YXc7Trat49Vyg4e5C1aYulIdH2AkgDiFf8tAJ7yP34KwMeDHWSMeQnAQJzXUsIRjWi0tjoTuIWFTkH3qSmuvj12zJkUni0c2JaCtAXRu7qYfA6gxZ6b63QANq4f4HUAir8tFWnZs4fH2vQRpaU8V1sbXT4bN4af+GxoYOfQ2MjXzp/PTmRkhKGd4Vwo9fWcgLZzBuXl7ETfeSd6108sLhxdBKUkgHijfaqNMR0AYIzpEJFF8ZxMRO4EcCcA1OsQN3qiWVRki5qXlDiraScmKLatrUzhYFNH27q3IhT70VGew+b6Lyyk+A8PcwHZli3MFnr8ON0+7vw9ubm0xk+dciqLubEW9uCgk+jNrjC+6ipGIj3ySGQFV+69lz7+7m5a/Db99YULoV0oW7cCn/600yGOjLD969ZFn5Qt1hQPughKmWPCWv4i8qKIHApy2+J1Y4wxTxhjGo0xjT6fz+vTZweR1nndutXJw2Nj6ScmHLG/5hqKu13gZSeIbRI4gK/Ny2No5sKFdOWMjVG8V6zgc8Ap/mJfYyuStbczHYWluZnWdX8/5wlsoZeFCymWb7wRue/chrxWVvJ91tZyfiMSF0pDAwveVFQ4bbn6arqiop10VReOkqKEtfyNMTfOtk9EukRksd/qXwyg29PWKXNHQwNTQ3zlK04qZhEK3Y03Uiyvu47ZP+2KXlvA3ebrsRZ5ays7EJ+PotndTdfR+DhfY0cHdtVwb6+T2uG551iJrKGBYr1+PecNbB1hW5vAhn/OlmjNvb6hsJCT1ytXcl9pKSONhoY4GR3JYq0NG4LnnI92RBpqNKYLuZQkEq/bZy+A2wA87L//UdwtUhJHYE6gwFQIixczDHPfPro+7MigqIgLuU6edEYNtbXsOFasoFgPDjo5fSor6V6yKZbHx3kOG7JpFzC1tlKwx8eZebS/3yk839vL0Ygb6zsPXBT1wgt8rS3QsmoVRw/RFAvxMvOkFXT7Oe/ZwzkVmz5CF3IpSSCu3D4isgDA9wHUA2gF8AljzHkRaQTwOWPMZ/3H/RrA5QDKAJwDcIcx5oVQ59bcPknGbZUWFFCE9+/n4/p6CnxnJ/33AwOciF27lsnXTp2ilS1C4R8ddfLot7YyYqe4mJ1GcTGzkbqLu9u00zbrZ0EBrffGRnZWlmBF4QFG50RTySuSzyAeyzxYRa99+ziHEOz9JKualZIRJCS3jzHmHIAbgmxvAvBZ1/MPxXMdxSMiFbPZEqMBwI9/zEiY6mqnmHxpKSduq6spsAUFFHyb96e/n6OG3l6eo7PTmQe49lrHgr/nHork0aPsNAoL2cmUldEt9cILnE/w+Xi9vDxa4o88Mr1oymyVvGJx2XhhhQer6DU+zkgnt/jrQi4lgegK32wh0sVGwY774hdZML2ggL733FyK/OnTTqhoUxOPBSjUNhHc2bPTE8KNjHDhVUkJz3XsGDNs1tc7/vHRUYp1dzeFv7+fojjgjxbu6WERls2b+ZrA9Q1r1vBYGy2U7EnWYOGePt/M1Nq6kEtJICr+2UKki42CHdfTQyGuqqKYLlnipHeuqXFcOLZgyuQkY+vdK4Jzc3krKeFcQE0NjxGhkFthbmhgBtDKSnYO5887Sd4Ajhquu47unEOHuC0woqaggJE5tpB8suPkgy2+s5+hRgEpSULFP1uIdLFRsOPsgiv7ms5OCqyNEBLhhG91NXDnnRSzd97hfhsdZEN3c3PprikupkVfUcEIIbcwb91Kwe/s5EjBCn9eHt05Bw4ET1vhXhT10EPAt78dPuQ1EQQL98zNZbSVLuRSkoSmdM4WIl1sFOw4d1rmNWvoGsrPdyz+kREuwPrVr5gEzudjZzE0xA5g3jx2Fvn53F5T40zEBl4LoADeeCMLrNu6AsZw3iAvj2IZLG1FqgpnqHDPUCmlFWUOUfHPFiINXQx2nM9H6/74cSclxMCAkzVz40aK2quvOuGdJSXsNPr6uCbAHeY5Pg788pd0feTmBg+f3L4d+O53nTKOY2NOVbHR0fQr+J3KnZOSlWgZx2winmifY8e4IMxm6bTF2a+9lhO/L79MUS4upkCPjbEDyMmhe6ekhK+9/HJ2KoE1iINd9+mnuQrYLg6zKSIWLgSef17FVFGCoGUclZlEan0GO27PHrpqrIumq4sZQH/3O4Zw1tUxdHF42En1bCeAfT5a+b299ON3dvL56tWctN22LXj1qvFxziNMTDgun9JSricI9z509ayihEQnfJXICJwINoaW/+CgsxI3N9cpAA9Q/HNzgb//e3YWNq1Dbi4Xgv3618Czz7LIyd13U+DdUUa1tZwvWLWKz234aG9v6OyaXhdjUZQMRMVfiQx3uGJnJ/37fX1OWgebpXP+fKcTyMsDPvABzhX09rKjsLV+Rej6GRigQHd3cxTQ2elcc8MGTibX1fHclZXsDGprQ4u5lkFUlLCo+CuR4Q5XPHLEybi5di0FfnKSwl5czBFCeTkzdo6NAS++yHOMj3Py99w5jhampijuzz3H146OAm+/7VyzqIhRP2fOOG6fggJe/+hRYNeu4G3VMoiKEhYVfyUy3LH07e0U02uuoU9+7VouqjKG1r3NDFpXR4vfTgLbcpETE+wUxsd5n5NDi76riwLtXvS0fTst/uJiJpLr6HAmf3/+8+DWv5ZBVJSw6ISvEjnuieDA+PwrrqBVvmnT9FDSykon1cLYGK33ycnpKaTb2oClS9mRjI7yuY2FB7hgrLOTxwLcb1M8Byuu4mVGTkXJUDTUU4meYFkqL1xgrp1Dh5h8rbeX4v7OO3TzFBbStWOrYgHsEBYu5HMRivmKFcAPf+hca+dOhpk+/zzFPy+P5xBhEfpDh7jdGC40275dc+UrWY2GeipzR6gVq6tX0z1TUkJhthb/6ChdPtb1MznpiHZ/P59XVXH1r5vWVrqUamq4nmBigh1JURFLRQ4MsMMAuMK4rQ346ldDh7VqxxAZ+jllNCr+SmzMJq420ubNNynu8+czqufiRecYOwdgDDuG0lLHHXTmDEXHnru+npa/FX1bday/nyOApUvZ0QBOBFGoOrvB1hNoEZWZ6OeU8eiEr+ItNtKmr4/WOUDRLi+nn97W/V2yxKkNPD7OqmGbNtGKd4dkbt3KdNF9fewsbPGYqSkuKBsYcDqWoiKOMEJF9WgYaGTo55TxqOWveItNDFdR4VTqsumca2qYwXPNGq4ONobpHtascdw9U1MzxXt4mK8XoRCNjrJDuXiR+9ranDxBhYWho3paW6cXfgE0DDQY+jllPCr+irfYSJvaWvr8R0fplpmYoKtm6VIK/xwMWdwAAAmcSURBVLlztPLr6qb7+QNDMvfs4bkAZzGZDfUsKXHSRnd0cHSxcqWTEz+YzzrS7KbZjn5OGY+6fRRvsZPBq1dT3CsrWav2lluA9esZCQQAN9wAXHYZ6wIfO0aL//hx1rY9eJBRPs3NFG670temj7ZrBW64AVi+nOcbHmaRFzvZO1uKh/XrZ+bW1yIqMwlWg0A/p4wi3gLu8wF8D8ByAKcA/HdjzIWAYzYA2AWgHMAkgIeMMd8Ld24N9cxAdu6caU0eP85J3spKhoWuW8foHhs+Wlrq1AR++23ejOEI4qMf5TmCFT4Pdi37fOtWjWKJBI32SUsSFep5H4CXjDEPi8h9/uf/EHDMEIC/NsYcF5FaAK+JyAvGmN44r62kG8H8yCtXOn76ZcumFzkHnNz9VVVMH11by9HCZZfRIp1tAVcon7Xm1o8M/ZwymnjdPlsAPOV//BSAjwceYIw5Zow57n/cDqAbgC/O6yrpSKi0C7Pl4xkbm16icfVq4OtfZ6bPUOUPw6V4aG7m6OD22x0Xk6JkEfGKf7UxpgMA/PeLQh0sIlcCKABwYpb9d4pIk4g09fT0xNk0JeUI5UcOJdYNDRToJ5/ksYcOhXdFhLqWpnxWlPDiLyIvisihILct0VxIRBYD+A6A/2GMmQp2jDHmCWNMozGm0efTwUHGEazQurXaI5lgjEa0Q11LY9gVJbzP3xhz42z7RKRLRBYbYzr84t49y3HlAH4C4AFjzP6YW6ukP7P5kUOljLC4RRtw7gNX9AZOVN5zz/T9GsOuKHFP+O4FcBuAh/33Pwo8QEQKAPwAwNPGmP+K83pKJhNugjES0Y4kLYHGsCtK3D7/hwHcJCLHAdzkfw4RaRSRf/Mf898BXAvgMyJy0H/bEOd1lWwkkjz9kbh0NIZdUeITf2PMOWPMDcaYVf778/7tTcaYz/of/4cxJt8Ys8F1O+hF45UsIxLRjqSKV6j5AEXJEjS9g5I+RDIvEKlLJ9DFZEM/dUGTkiWo+CvpRbh5gViqeDU3A/ffzyLyo6PA4cOsFfDQQ9oBKBmL5vZRMotYXDq7dgEtLXxsXUYtLbMXiFeUDEAtfyXziDYtwf79LBJjawTbCmP7NSpZyVzU8leU2ZIbpmh9a0XxAhV/RbnqKlYEGx6m4NsKYVddleyWKcqcoW4fJX2YqxTD27dzfqCnhxPEhYXMNrp9e/znVpQUJa58/nOJ5vNXpuFeueuO4vEqPl9z1ysZQqLy+StKYog0r0+saO56JctQn7+SHkSycldRlIhR8VfSg0jy+iiKEjEq/kp6oMnYFMVTVPyV9ECTsSmKp+iEr5I+6KSsoniGWv6KoihZiIq/oihKFqLiryiKkoWo+CuKomQhKv6KoihZSFziLyLzReTnInLcf18V5JhlIvKav3D7YRH5XDzXVBRFUeInXsv/PgAvGWNWAXjJ/zyQDgDXGGM2AHg/gPtEpDbO6yqKoihxEK/4bwHwlP/xUwA+HniAMWbMGDPqf1rowTUVRVGUOIlXiKuNMR0A4L9fFOwgEVkqIs0ATgP4mjGmfZbj7hSRJhFp6unpibNpiqIoymyEzecvIi8CqAmy634ATxljKl3HXjDGzPD7u/bXAvghgI8ZY7rCXLcHwLshGxcfCwGcncPze0U6tFPb6A3aRm/I9jYuM8b4wh0UNr2DMebG2faJSJeILDbGdIjIYgDdYc7VLiKHAXwIwO4wx4ZtfDyISFMkBQ+STTq0U9voDdpGb9A2Rka8bp+9AG7zP74NwI8CDxCRJSJS7H9cBeADAI7GeV1FURQlDuIV/4cB3CQixwHc5H8OEWkUkX/zH7MGwO9E5E0AvwLwDWPMH+K8rqIoihIHcWX1NMacA3BDkO1NAD7rf/xzAKmYivGJZDcgQtKhndpGb9A2eoO2MQJStoC7oiiKMndozL2iKEoWouKvKIqShWSN+EeYh2iDiLzqz0HULCKfTMV2+o97XkR6ReTHCWrXLSJyVERaRGRGGg8RKRSR7/n3/05ElieiXVG28VoReV1EJkRkW6Lb52pHuHb+rYi85f8NviQiy1KwjZ8TkT/4c3b9RkTWplobXcdtExEjIgkPrYzgc/yMiPT4P8eDIvLZhDXOGJMVNwBfB3Cf//F94ErjwGNWA1jlf1wL5iWqTLV2+vfdAOBjAH6cgDblAjgBYAWAAgBvAlgbcMx2AN/2P74VwPcS/LlF0sblYPDB0wC2Jel3GEk7rwdQ4n98V4p+luWux5sBPJ9qbfQfNw/AywD2A2hMtTYC+AyAR5PxW8wayx+R5SE6Zow57n/cDi5am9PFZkEI204AMMa8BGAgQW26EkCLMeakMWYMwLNgO924270bwA0iIglqX0RtNMacMsY0A5hKYLsCiaSdvzTGDPmf7gewJAXb2O96Wgog0ZEjkfwmAeAroEE1ksjG+Ym0jUkhm8Q/ojxEFhG5EuytTySgbW6iameCqAPzMlna/NuCHmOMmQDQB2BBQloXcH0/wdqYCkTbzjsA/GxOWzSTiNooIn8jIidAcb07QW2zhG2jiGwEsNQYkxDXaBAi/a7/wu/i2y0iSxPTtDjj/FONMHmIojnPYgDfAXCbMcZzK9GrdiaQYBZ8oKUXyTFzSbKvHykRt1NEPgWgEcB1c9qiIJcOsm1GG40xjwF4TET+CsADcFb7J4KQbRSRHADfBN0qySKSz/E5AN81xoz6a508BeBP57xlyDDxNx7kIRKRcgA/AfCAMWZ/qrYzwbQBcFskSwAEZma1x7SJSB6ACgDnE9O8ade3BGtjKhBRO0XkRtAYuM44KdETRbSf5bMAds1pi2YSro3zAKwHsM/vfawBsFdENhsuQk2FNsJwoazlXwF8LQHtApBdbp9I8hAVAPgBgKeNMf+VwLa5CdvOJHAAwCoRucT/Gd0KttONu93bAPzC+Ge0UqiNqUDYdvrdFf8CYLMxJhmdfyRtXOV6+hEAxxPYPiBMG40xfcaYhcaY5caY5eDcSSKFP2wbgT96GSybARxJWOuSMcucjBvof34J/JG+BGC+f3sjgH/zP/4UgHEAB123DanWTv/zXwPoATAMWhg3z3G7/gzAMXAO5H7/tgfBPxQAFAH4LwAtAH4PYEUSvuNwbXyf/7MaBHAOwOEk/RbDtfNFAF2u3+DeFGzjPwM47G/fLwGsS7U2Bhy7DwmO9onwc/y//Z/jm/7P8fJEtU3TOyiKomQh2eT2URRFUfyo+CuKomQhKv6KoihZiIq/oihKFqLiryiKkoWo+CuKomQhKv6KoihZyP8POCH37sFBNGYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data2D = PCA(n_components = 2).fit_transform(bernardo_tfidf.todense())\n",
    "plt.scatter(data2D[:, 0], data2D[:, 1], c = 'red', alpha = 0.5)\n",
    "plt.title('Bernardo Soares')\n",
    "plt.show()\n",
    "\n",
    "data2D = PCA(n_components = 2).fit_transform(alberto_tfidf.todense())\n",
    "plt.scatter(data2D[:, 0], data2D[:, 1], c = 'purple', alpha = 0.5)\n",
    "plt.title('Alberto Caeiro')\n",
    "plt.show()\n",
    "\n",
    "data2D = PCA(n_components = 2).fit_transform(ricardo_tfidf.todense())\n",
    "plt.scatter(data2D[:, 0], data2D[:, 1], c = 'blue', alpha = 0.5)\n",
    "plt.title('Ricardo Reis')\n",
    "plt.show()\n",
    "\n",
    "data2D = PCA(n_components = 2).fit_transform(campos_tfidf.todense())\n",
    "plt.scatter(data2D[:, 0], data2D[:, 1], c = 'green', alpha = 0.5)\n",
    "plt.title('Álvaro de Campos')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "pessoa.ipynb",
   "provenance": []
  },
  "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}