08. 감성분석.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8.2 감성 사전을 이용한 영화 리뷰 감성 분석\n",
    "- 어휘 기반\n",
    "### 8.2.1 NLTK 영화 리뷰 데이터 준비"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "d:\\anaconda3\\lib\\site-packages\\numpy\\_distributor_init.py:30: UserWarning: loaded more than 1 DLL from .libs:\n",
      "d:\\anaconda3\\lib\\site-packages\\numpy\\.libs\\libopenblas.4SP5SUA7CBGXUEOC35YP2ASOICYYEQZZ.gfortran-win_amd64.dll\n",
      "d:\\anaconda3\\lib\\site-packages\\numpy\\.libs\\libopenblas64__v0.3.23-gcc_10_3_0.dll\n",
      "  warnings.warn(\"loaded more than 1 DLL from .libs:\"\n",
      "[nltk_data] Downloading package movie_reviews to\n",
      "[nltk_data]     C:\\Users\\namyo\\AppData\\Roaming\\nltk_data...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#review count : 2000\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data]   Package movie_reviews is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "import nltk\n",
    "nltk.download('movie_reviews')\n",
    "\n",
    "from nltk.corpus import movie_reviews\n",
    "print(\"#review count : {}\".format(len(movie_reviews.fileids())))  # 영화 리뷰 문서의 id를 반환"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#samples of file ids : ['neg/cv000_29416.txt', 'neg/cv001_19502.txt', 'neg/cv002_17424.txt', 'neg/cv003_12683.txt', 'neg/cv004_12641.txt', 'neg/cv005_29357.txt', 'neg/cv006_17022.txt', 'neg/cv007_4992.txt', 'neg/cv008_29326.txt', 'neg/cv009_29417.txt']\n",
      "#categories of reviews : ['neg', 'pos']\n",
      "#num of neg reviews : 1000\n",
      "#num of pos reviews : 1000\n",
      "#id of the first review : neg/cv000_29416.txt\n",
      "part of the first review : plot : two teen couples go to a church party , drink and then drive . \n",
      "they get into an accident . \n",
      "one of the guys dies , but his girlfriend continues to see him in her life , and has nightmares . \n",
      "what's the deal ? \n",
      "watch the movie and \" sorta \" find out . . . \n",
      "critique : a mind-fuck movie for the teen generation that touches on a very cool idea , but presents it in a very bad package . \n",
      "which is what makes this review an even harder one to write , since i generally applaud films which attempt\n",
      "sentiment of the first review : ['neg']\n"
     ]
    }
   ],
   "source": [
    "print('#samples of file ids : {}'.format(movie_reviews.fileids()[:10]))  # id를 10개까지만 출력\n",
    "print('#categories of reviews : {}'.format(movie_reviews.categories()))  # label\n",
    "print('#num of neg reviews : {}'.format(len(movie_reviews.fileids(categories='neg'))))\n",
    "print('#num of pos reviews : {}'.format(len(movie_reviews.fileids(categories='pos'))))\n",
    "\n",
    "fileid = movie_reviews.fileids()[0]  # 첫 번째 문서 id를 반환\n",
    "print('#id of the first review : {}'.format(fileid))\n",
    "print('#part of the first review : {}'.format(movie_reviews.raw(fileid)[:500]))\n",
    "print('#sentiment of the first review : {}'.format(movie_reviews.categories(fileid)))\n",
    "\n",
    "# 영화 리뷰 데이터에서 파일 id를 가져옴\n",
    "fileids = movie_reviews.fileids()\n",
    "\n",
    "# 파일 id를 이용해 raw text file을 가져옴\n",
    "reviews = [ movie_reviews.raw(fileid) for fileid in fileids ]\n",
    "categories = [ movie_reviews.categories(fileid)[0] for fileid in fileids ]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### [8.2.2 TextBlob을 이용한 감성 분석](https://textblob.readthedocs.io/en/dev/)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Part of speech tagging : [('plot', 'NN'), ('two', 'CD'), ('teen', 'NN'), ('couples', 'NNS'), ('go', 'VBP'), ('to', 'TO'), ('a', 'DT'), ('church', 'NN'), ('party', 'NN'), ('drink', 'NN'), ('and', 'CC'), ('then', 'RB'), ('drive', 'NN'), ('they', 'PRP'), ('get', 'VBP'), ('into', 'IN'), ('an', 'DT'), ('accident', 'NN'), ('one', 'CD'), ('of', 'IN'), ('the', 'DT'), ('guys', 'NN'), ('dies', 'VBZ'), ('but', 'CC'), ('his', 'PRP$'), ('girlfriend', 'NN'), ('continues', 'VBZ'), ('to', 'TO'), ('see', 'VB'), ('him', 'PRP'), ('in', 'IN'), ('her', 'PRP$'), ('life', 'NN'), ('and', 'CC'), ('has', 'VBZ'), ('nightmares', 'NNS'), ('what', 'WP'), (\"'s\", 'VBZ'), ('the', 'DT'), ('deal', 'NN'), ('watch', 'VB'), ('the', 'DT'), ('movie', 'NN'), ('and', 'CC'), ('sorta', 'JJ'), ('find', 'VB'), ('out', 'RP'), ('critique', 'NN'), ('a', 'DT'), ('mind-fuck', 'JJ'), ('movie', 'NN'), ('for', 'IN'), ('the', 'DT'), ('teen', 'JJ'), ('generation', 'NN'), ('that', 'WDT'), ('touches', 'VBZ'), ('on', 'IN'), ('a', 'DT'), ('very', 'RB'), ('cool', 'JJ'), ('idea', 'NN'), ('but', 'CC'), ('presents', 'VBZ'), ('it', 'PRP'), ('in', 'IN'), ('a', 'DT'), ('very', 'RB'), ('bad', 'JJ'), ('package', 'NN'), ('which', 'WDT'), ('is', 'VBZ'), ('what', 'WP'), ('makes', 'VBZ'), ('this', 'DT'), ('review', 'NN'), ('an', 'DT'), ('even', 'RB'), ('harder', 'RBR'), ('one', 'CD'), ('to', 'TO'), ('write', 'VB'), ('since', 'IN'), ('i', 'NN'), ('generally', 'RB'), ('applaud', 'VBP'), ('films', 'NNS'), ('which', 'WDT'), ('attempt', 'VBP'), ('to', 'TO'), ('break', 'VB'), ('the', 'DT'), ('mold', 'NN'), ('mess', 'NN'), ('with', 'IN'), ('your', 'PRP$'), ('head', 'NN'), ('and', 'CC'), ('such', 'JJ'), ('lost', 'VBN'), ('highway', 'NN'), ('&', 'CC'), ('memento', 'NN'), ('but', 'CC'), ('there', 'EX'), ('are', 'VBP'), ('good', 'JJ'), ('and', 'CC'), ('bad', 'JJ'), ('ways', 'NNS'), ('of', 'IN'), ('making', 'VBG'), ('all', 'DT'), ('types', 'NNS'), ('of', 'IN'), ('films', 'NNS'), ('and', 'CC'), ('these', 'DT'), ('folks', 'NNS'), ('just', 'RB'), ('did', 'VBD'), (\"n't\", 'RB'), ('snag', 'VB'), ('this', 'DT'), ('one', 'CD'), ('correctly', 'RB'), ('they', 'PRP'), ('seem', 'VBP'), ('to', 'TO'), ('have', 'VB'), ('taken', 'VBN'), ('this', 'DT'), ('pretty', 'RB'), ('neat', 'JJ'), ('concept', 'NN'), ('but', 'CC'), ('executed', 'VBD'), ('it', 'PRP'), ('terribly', 'RB'), ('so', 'RB'), ('what', 'WP'), ('are', 'VBP'), ('the', 'DT'), ('problems', 'NNS'), ('with', 'IN'), ('the', 'DT'), ('movie', 'NN'), ('well', 'RB'), ('its', 'PRP$'), ('main', 'JJ'), ('problem', 'NN'), ('is', 'VBZ'), ('that', 'IN'), ('it', 'PRP'), (\"'s\", 'VBZ'), ('simply', 'RB'), ('too', 'RB'), ('jumbled', 'VBN'), ('it', 'PRP'), ('starts', 'VBZ'), ('off', 'IN'), ('normal', 'JJ'), ('but', 'CC'), ('then', 'RB'), ('downshifts', 'VBZ'), ('into', 'IN'), ('this', 'DT'), ('fantasy', 'JJ'), ('world', 'NN'), ('in', 'IN'), ('which', 'WDT'), ('you', 'PRP'), ('as', 'IN'), ('an', 'DT'), ('audience', 'NN'), ('member', 'NN'), ('have', 'VBP'), ('no', 'DT'), ('idea', 'NN'), ('what', 'WP'), (\"'s\", 'VBZ'), ('going', 'VBG'), ('on', 'IN'), ('there', 'EX'), ('are', 'VBP'), ('dreams', 'NNS'), ('there', 'EX'), ('are', 'VBP'), ('characters', 'NNS'), ('coming', 'VBG'), ('back', 'RB'), ('from', 'IN'), ('the', 'DT'), ('dead', 'JJ'), ('there', 'EX'), ('are', 'VBP'), ('others', 'NNS'), ('who', 'WP'), ('look', 'VBP'), ('like', 'IN'), ('the', 'DT'), ('dead', 'JJ'), ('there', 'EX'), ('are', 'VBP'), ('strange', 'JJ'), ('apparitions', 'NNS'), ('there', 'EX'), ('are', 'VBP'), ('disappearances', 'NNS'), ('there', 'EX'), ('are', 'VBP'), ('a', 'DT'), ('looooot', 'NN'), ('of', 'IN'), ('chase', 'NN'), ('scenes', 'NNS'), ('there', 'EX'), ('are', 'VBP'), ('tons', 'NNS'), ('of', 'IN'), ('weird', 'JJ'), ('things', 'NNS'), ('that', 'WDT'), ('happen', 'VBP'), ('and', 'CC'), ('most', 'JJS'), ('of', 'IN'), ('it', 'PRP'), ('is', 'VBZ'), ('simply', 'RB'), ('not', 'RB'), ('explained', 'VBN'), ('now', 'RB'), ('i', 'VBZ'), ('personally', 'RB'), ('do', 'VBP'), (\"n't\", 'RB'), ('mind', 'VB'), ('trying', 'VBG'), ('to', 'TO'), ('unravel', 'VB'), ('a', 'DT'), ('film', 'NN'), ('every', 'DT'), ('now', 'RB'), ('and', 'CC'), ('then', 'RB'), ('but', 'CC'), ('when', 'WRB'), ('all', 'DT'), ('it', 'PRP'), ('does', 'VBZ'), ('is', 'VBZ'), ('give', 'VB'), ('me', 'PRP'), ('the', 'DT'), ('same', 'JJ'), ('clue', 'NN'), ('over', 'IN'), ('and', 'CC'), ('over', 'RB'), ('again', 'RB'), ('i', 'JJ'), ('get', 'VBP'), ('kind', 'NN'), ('of', 'IN'), ('fed', 'VBN'), ('up', 'RP'), ('after', 'IN'), ('a', 'DT'), ('while', 'NN'), ('which', 'WDT'), ('is', 'VBZ'), ('this', 'DT'), ('film', 'NN'), (\"'s\", 'POS'), ('biggest', 'JJS'), ('problem', 'NN'), ('it', 'PRP'), (\"'s\", 'VBZ'), ('obviously', 'RB'), ('got', 'VBD'), ('this', 'DT'), ('big', 'JJ'), ('secret', 'NN'), ('to', 'TO'), ('hide', 'VB'), ('but', 'CC'), ('it', 'PRP'), ('seems', 'VBZ'), ('to', 'TO'), ('want', 'VB'), ('to', 'TO'), ('hide', 'VB'), ('it', 'PRP'), ('completely', 'RB'), ('until', 'IN'), ('its', 'PRP$'), ('final', 'JJ'), ('five', 'CD'), ('minutes', 'NNS'), ('and', 'CC'), ('do', 'VBP'), ('they', 'PRP'), ('make', 'VB'), ('things', 'NNS'), ('entertaining', 'VBG'), ('thrilling', 'VBG'), ('or', 'CC'), ('even', 'RB'), ('engaging', 'VBG'), ('in', 'IN'), ('the', 'DT'), ('meantime', 'NN'), ('not', 'RB'), ('really', 'RB'), ('the', 'DT'), ('sad', 'JJ'), ('part', 'NN'), ('is', 'VBZ'), ('that', 'IN'), ('the', 'DT'), ('arrow', 'NN'), ('and', 'CC'), ('i', 'VB'), ('both', 'DT'), ('dig', 'NN'), ('on', 'IN'), ('flicks', 'NNS'), ('like', 'IN'), ('this', 'DT'), ('so', 'IN'), ('we', 'PRP'), ('actually', 'RB'), ('figured', 'VBD'), ('most', 'JJS'), ('of', 'IN'), ('it', 'PRP'), ('out', 'RP'), ('by', 'IN'), ('the', 'DT'), ('half-way', 'JJ'), ('point', 'NN'), ('so', 'IN'), ('all', 'DT'), ('of', 'IN'), ('the', 'DT'), ('strangeness', 'NN'), ('after', 'IN'), ('that', 'DT'), ('did', 'VBD'), ('start', 'VB'), ('to', 'TO'), ('make', 'VB'), ('a', 'DT'), ('little', 'JJ'), ('bit', 'NN'), ('of', 'IN'), ('sense', 'NN'), ('but', 'CC'), ('it', 'PRP'), ('still', 'RB'), ('did', 'VBD'), (\"n't\", 'RB'), ('the', 'DT'), ('make', 'VB'), ('the', 'DT'), ('film', 'NN'), ('all', 'PDT'), ('that', 'IN'), ('more', 'JJR'), ('entertaining', 'NN'), ('i', 'NN'), ('guess', 'VBP'), ('the', 'DT'), ('bottom', 'JJ'), ('line', 'NN'), ('with', 'IN'), ('movies', 'NNS'), ('like', 'IN'), ('this', 'DT'), ('is', 'VBZ'), ('that', 'IN'), ('you', 'PRP'), ('should', 'MD'), ('always', 'RB'), ('make', 'VB'), ('sure', 'JJ'), ('that', 'IN'), ('the', 'DT'), ('audience', 'NN'), ('is', 'VBZ'), ('into', 'IN'), ('it', 'PRP'), ('even', 'RB'), ('before', 'IN'), ('they', 'PRP'), ('are', 'VBP'), ('given', 'VBN'), ('the', 'DT'), ('secret', 'JJ'), ('password', 'NN'), ('to', 'TO'), ('enter', 'VB'), ('your', 'PRP$'), ('world', 'NN'), ('of', 'IN'), ('understanding', 'NN'), ('i', 'JJ'), ('mean', 'VBP'), ('showing', 'VBG'), ('melissa', 'JJ'), ('sagemiller', 'NN'), ('running', 'VBG'), ('away', 'RB'), ('from', 'IN'), ('visions', 'NNS'), ('for', 'IN'), ('about', 'IN'), ('20', 'CD'), ('minutes', 'NNS'), ('throughout', 'IN'), ('the', 'DT'), ('movie', 'NN'), ('is', 'VBZ'), ('just', 'RB'), ('plain', 'JJ'), ('lazy', 'JJ'), ('okay', 'RB'), ('we', 'PRP'), ('get', 'VBP'), ('it', 'PRP'), ('there', 'EX'), ('are', 'VBP'), ('people', 'NNS'), ('chasing', 'VBG'), ('her', 'PRP$'), ('and', 'CC'), ('we', 'PRP'), ('do', 'VBP'), (\"n't\", 'RB'), ('know', 'VB'), ('who', 'WP'), ('they', 'PRP'), ('are', 'VBP'), ('do', 'VBP'), ('we', 'PRP'), ('really', 'RB'), ('need', 'VB'), ('to', 'TO'), ('see', 'VB'), ('it', 'PRP'), ('over', 'IN'), ('and', 'CC'), ('over', 'RB'), ('again', 'RB'), ('how', 'WRB'), ('about', 'IN'), ('giving', 'VBG'), ('us', 'PRP'), ('different', 'JJ'), ('scenes', 'NNS'), ('offering', 'VBG'), ('further', 'JJ'), ('insight', 'NN'), ('into', 'IN'), ('all', 'DT'), ('of', 'IN'), ('the', 'DT'), ('strangeness', 'NN'), ('going', 'VBG'), ('down', 'RP'), ('in', 'IN'), ('the', 'DT'), ('movie', 'NN'), ('apparently', 'RB'), ('the', 'DT'), ('studio', 'NN'), ('took', 'VBD'), ('this', 'DT'), ('film', 'NN'), ('away', 'RB'), ('from', 'IN'), ('its', 'PRP$'), ('director', 'NN'), ('and', 'CC'), ('chopped', 'VBD'), ('it', 'PRP'), ('up', 'RP'), ('themselves', 'PRP'), ('and', 'CC'), ('it', 'PRP'), ('shows', 'VBZ'), ('there', 'RB'), ('might', 'MD'), (\"'ve\", 'VBP'), ('been', 'VBN'), ('a', 'DT'), ('pretty', 'JJ'), ('decent', 'JJ'), ('teen', 'JJ'), ('mind-fuck', 'JJ'), ('movie', 'NN'), ('in', 'IN'), ('here', 'RB'), ('somewhere', 'RB'), ('but', 'CC'), ('i', 'VBZ'), ('guess', 'VBP'), ('the', 'DT'), ('suits', 'NNS'), ('decided', 'VBD'), ('that', 'IN'), ('turning', 'VBG'), ('it', 'PRP'), ('into', 'IN'), ('a', 'DT'), ('music', 'NN'), ('video', 'NN'), ('with', 'IN'), ('little', 'JJ'), ('edge', 'NN'), ('would', 'MD'), ('make', 'VB'), ('more', 'JJR'), ('sense', 'NN'), ('the', 'DT'), ('actors', 'NNS'), ('are', 'VBP'), ('pretty', 'RB'), ('good', 'JJ'), ('for', 'IN'), ('the', 'DT'), ('most', 'JJS'), ('part', 'NN'), ('although', 'IN'), ('wes', 'JJ'), ('bentley', 'NN'), ('just', 'RB'), ('seemed', 'VBD'), ('to', 'TO'), ('be', 'VB'), ('playing', 'VBG'), ('the', 'DT'), ('exact', 'JJ'), ('same', 'JJ'), ('character', 'NN'), ('that', 'IN'), ('he', 'PRP'), ('did', 'VBD'), ('in', 'IN'), ('american', 'JJ'), ('beauty', 'NN'), ('only', 'RB'), ('in', 'IN'), ('a', 'DT'), ('new', 'JJ'), ('neighborhood', 'NN'), ('but', 'CC'), ('my', 'PRP$'), ('biggest', 'JJS'), ('kudos', 'NN'), ('go', 'VBP'), ('out', 'RP'), ('to', 'TO'), ('sagemiller', 'VB'), ('who', 'WP'), ('holds', 'VBZ'), ('her', 'PRP$'), ('own', 'JJ'), ('throughout', 'IN'), ('the', 'DT'), ('entire', 'JJ'), ('film', 'NN'), ('and', 'CC'), ('actually', 'RB'), ('has', 'VBZ'), ('you', 'PRP'), ('feeling', 'VBG'), ('her', 'PRP$'), ('character', 'NN'), (\"'s\", 'POS'), ('unraveling', 'NN'), ('overall', 'JJ'), ('the', 'DT'), ('film', 'NN'), ('does', 'VBZ'), (\"n't\", 'RB'), ('stick', 'VB'), ('because', 'IN'), ('it', 'PRP'), ('does', 'VBZ'), (\"n't\", 'RB'), ('entertain', 'VB'), ('it', 'PRP'), (\"'s\", 'VBZ'), ('confusing', 'VBG'), ('it', 'PRP'), ('rarely', 'RB'), ('excites', 'VBZ'), ('and', 'CC'), ('it', 'PRP'), ('feels', 'VBZ'), ('pretty', 'JJ'), ('redundant', 'NN'), ('for', 'IN'), ('most', 'JJS'), ('of', 'IN'), ('its', 'PRP$'), ('runtime', 'NN'), ('despite', 'IN'), ('a', 'DT'), ('pretty', 'JJ'), ('cool', 'NN'), ('ending', 'VBG'), ('and', 'CC'), ('explanation', 'NN'), ('to', 'TO'), ('all', 'DT'), ('of', 'IN'), ('the', 'DT'), ('craziness', 'NN'), ('that', 'WDT'), ('came', 'VBD'), ('before', 'IN'), ('it', 'PRP'), ('oh', 'UH'), ('and', 'CC'), ('by', 'IN'), ('the', 'DT'), ('way', 'NN'), ('this', 'DT'), ('is', 'VBZ'), ('not', 'RB'), ('a', 'DT'), ('horror', 'NN'), ('or', 'CC'), ('teen', 'JJ'), ('slasher', 'JJR'), ('flick', 'NN'), ('it', 'PRP'), (\"'s\", 'VBZ'), ('just', 'RB'), ('packaged', 'VBN'), ('to', 'TO'), ('look', 'VB'), ('that', 'DT'), ('way', 'NN'), ('because', 'IN'), ('someone', 'NN'), ('is', 'VBZ'), ('apparently', 'RB'), ('assuming', 'VBG'), ('that', 'IN'), ('the', 'DT'), ('genre', 'NN'), ('is', 'VBZ'), ('still', 'RB'), ('hot', 'JJ'), ('with', 'IN'), ('the', 'DT'), ('kids', 'NNS'), ('it', 'PRP'), ('also', 'RB'), ('wrapped', 'VBD'), ('production', 'NN'), ('two', 'CD'), ('years', 'NNS'), ('ago', 'RB'), ('and', 'CC'), ('has', 'VBZ'), ('been', 'VBN'), ('sitting', 'VBG'), ('on', 'IN'), ('the', 'DT'), ('shelves', 'NNS'), ('ever', 'RB'), ('since', 'IN'), ('whatever', 'WDT'), ('skip', 'VB'), ('it', 'PRP'), ('where', 'WRB'), (\"'s\", 'POS'), ('joblo', 'NN'), ('coming', 'VBG'), ('from', 'IN'), ('a', 'DT'), ('nightmare', 'NN'), ('of', 'IN'), ('elm', 'JJ'), ('street', 'NN'), ('3', 'CD'), ('7/10', 'CD'), ('blair', 'NN'), ('witch', 'NN'), ('2', 'CD'), ('7/10', 'CD'), ('the', 'DT'), ('crow', 'NN'), ('9/10', 'CD'), ('the', 'DT'), ('crow', 'NN'), ('salvation', 'NN'), ('4/10', 'CD'), ('lost', 'VBN'), ('highway', 'NN'), ('10/10', 'CD'), ('memento', 'NN'), ('10/10', 'CD'), ('the', 'DT'), ('others', 'NNS'), ('9/10', 'CD'), ('stir', 'NN'), ('of', 'IN'), ('echoes', 'NNS'), ('8/10', 'CD')]\n",
      "Noun Phrase Extraction : ['teen couples', 'church party', 'mind-fuck movie', 'teen generation', 'cool idea', 'bad package', 'bad ways', \"n't snag\", 'pretty neat concept', 'main problem', 'audience member', 'strange apparitions', 'chase scenes', 'weird things', \"n't mind\", \"film 's\", 'sad part', 'half-way point', 'i guess', 'bottom line', 'secret password', 'melissa sagemiller', 'different scenes offering', 'pretty decent teen mind-fuck movie', 'i guess', 'music video', 'wes bentley', 'american beauty', 'new neighborhood', 'entire film', \"character 's\", 'teen slasher flick', \"'s joblo\", 'elm street', 'blair witch']\n",
      "Sentiment Analysis : Sentiment(polarity=0.06479782948532947, subjectivity=0.5188408350908352)\n"
     ]
    }
   ],
   "source": [
    "from textblob import TextBlob\n",
    "result = TextBlob(reviews[0])  # 첫째 리뷰에 대한 결과\n",
    "print(\"Part of speech tagging : {}\".format(result.tags, \"\\n\"))\n",
    "print(\"Noun Phrase Extraction : {}\".format(result.noun_phrases, \"\\n\"))\n",
    "print(\"Sentiment Analysis : {}\".format(result.sentiment))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sentiment_TextBlob(docs):\n",
    "    results = []\n",
    "    \n",
    "    for doc in docs:\n",
    "        testimonial = TextBlob(doc)\n",
    "        if testimonial.sentiment.polarity > 0:\n",
    "            results.append('pos')\n",
    "        else:\n",
    "            results.append('neg')\n",
    "    return results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#TextBlob을 이용한 리뷰 감성분석의 정확도 :  0.6\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "print('#TextBlob을 이용한 리뷰 감성분석의 정확도 : ',\n",
    "    accuracy_score(categories, sentiment_TextBlob(reviews))\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.2.3 AFINN을 이용한 감성 분석\n",
    "```python\n",
    "%pip install afinn\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#Afinn을 이용한 리뷰 감성 분석의 정확도 :  0.664\n"
     ]
    }
   ],
   "source": [
    "from afinn import Afinn\n",
    "\n",
    "def sentiment_Afinn(docs):\n",
    "    afn = Afinn(emoticons=True)\n",
    "    results = []\n",
    "    \n",
    "    for doc in docs:\n",
    "        if afn.score(doc) > 0:\n",
    "            results.append('pos')\n",
    "        else:\n",
    "            results.append('neg')\n",
    "    return results\n",
    "\n",
    "print('#Afinn을 이용한 리뷰 감성 분석의 정확도 : ',\n",
    "    accuracy_score(categories, sentiment_Afinn(reviews))\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.2.4 VADER를 이용한 감성 분석"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package vader_lexicon to\n",
      "[nltk_data]     C:\\Users\\namyo\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package vader_lexicon is already up-to-date!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import nltk\n",
    "nltk.download('vader_lexicon')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#VADER을 이용한 리뷰 감성 분석 정확도 :  0.635\n"
     ]
    }
   ],
   "source": [
    "from afinn import Afinn\n",
    "from nltk.sentiment.vader import SentimentIntensityAnalyzer\n",
    "\n",
    "def sentiment_vader(docs):\n",
    "    analyser = SentimentIntensityAnalyzer()\n",
    "    results = []\n",
    "    \n",
    "    for doc in docs:\n",
    "        score = analyser.polarity_scores(doc)\n",
    "        if score['compound'] > 0:\n",
    "            results.append('pos')\n",
    "        else:\n",
    "            results.append('neg')\n",
    "    \n",
    "    return results\n",
    "print('#VADER을 이용한 리뷰 감성 분석 정확도 : ',\n",
    "    accuracy_score(categories, sentiment_vader(reviews))\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8.3 학습을 통한 머신러닝 기반의 감성 분석\n",
    "### 8.3.1 NLTK 영화 리뷰에 대한 머신러닝 기반 감성 분석"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train set count : 1600\n",
      "Test set count : 400\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    reviews, categories, test_size=0.2, random_state=7\n",
    ")\n",
    "\n",
    "print('Train set count : {}'.format(len(X_train)))\n",
    "print('Test set count : {}'.format(len(X_test)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#Train set dimension : (1600, 36189)\n",
      "#Test set dimension : (400, 36189)\n",
      "#Train set score : 0.998\n",
      "#Test set score : 0.797\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "\n",
    "tfidf = TfidfVectorizer()\n",
    "\n",
    "X_train_tfidf = tfidf.fit_transform(X_train)\n",
    "X_test_tfidf = tfidf.transform(X_test)\n",
    "print('#Train set dimension : {}'.format(X_train_tfidf.shape))\n",
    "print('#Test set dimension : {}'.format(X_test_tfidf.shape))\n",
    "\n",
    "NB_clf = MultinomialNB(alpha=0.01)\n",
    "NB_clf.fit(X_train_tfidf, y_train)\n",
    "\n",
    "print('#Train set score : {:.3f}'.format(NB_clf.score(X_train_tfidf, y_train)))\n",
    "print('#Test set score : {:.3f}'.format(NB_clf.score(X_test_tfidf, y_test)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.3.2 다음 영화 리뷰에 대한 머신러닝 기반 감성 분석"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>review</th>\n",
       "      <th>rating</th>\n",
       "      <th>date</th>\n",
       "      <th>title</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>돈 들인건 티가 나지만 보는 내내 하품만</td>\n",
       "      <td>1</td>\n",
       "      <td>2018.10.29</td>\n",
       "      <td>인피니티 워</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>몰입할수밖에 없다. 어렵게 생각할 필요없다. 내가 전투에 참여한듯 손에 땀이남.</td>\n",
       "      <td>10</td>\n",
       "      <td>2018.10.26</td>\n",
       "      <td>인피니티 워</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>이전 작품에 비해 더 화려하고 스케일도 커졌지만.... 전국 맛집의 음식들을 한데 ...</td>\n",
       "      <td>8</td>\n",
       "      <td>2018.10.24</td>\n",
       "      <td>인피니티 워</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>이 정도면 볼만하다고 할 수 있음!</td>\n",
       "      <td>8</td>\n",
       "      <td>2018.10.22</td>\n",
       "      <td>인피니티 워</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>재미있다</td>\n",
       "      <td>10</td>\n",
       "      <td>2018.10.20</td>\n",
       "      <td>인피니티 워</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              review  rating        date  \\\n",
       "0                             돈 들인건 티가 나지만 보는 내내 하품만       1  2018.10.29   \n",
       "1       몰입할수밖에 없다. 어렵게 생각할 필요없다. 내가 전투에 참여한듯 손에 땀이남.      10  2018.10.26   \n",
       "2  이전 작품에 비해 더 화려하고 스케일도 커졌지만.... 전국 맛집의 음식들을 한데 ...       8  2018.10.24   \n",
       "3                                이 정도면 볼만하다고 할 수 있음!       8  2018.10.22   \n",
       "4                                               재미있다      10  2018.10.20   \n",
       "\n",
       "    title  \n",
       "0  인피니티 워  \n",
       "1  인피니티 워  \n",
       "2  인피니티 워  \n",
       "3  인피니티 워  \n",
       "4  인피니티 워  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "df = pd.read_csv('./data/daum_movie_review.csv')\n",
    "df.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.style.use('ggplot')\n",
    "\n",
    "df['rating'].value_counts().sort_index().plot(kind='bar')\n",
    "#df['rating'].plot.hist(bins=10)  # 히스토그램\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#Train set size : 11043\n",
      "#Test set size : 3682\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "d:\\anaconda3\\lib\\site-packages\\sklearn\\feature_extraction\\text.py:528: UserWarning: The parameter 'token_pattern' will not be used since 'tokenizer' is not None'\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#Regression Train set R2 score : 0.6045480181543007\n",
      "#Regression Test set R2 score : 0.39500332535456006\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "from konlpy.tag import Okt\n",
    "from sklearn.linear_model import LinearRegression\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    df['review'], df['rating'], random_state=7\n",
    ")\n",
    "print('#Train set size : {}'.format(len(X_train)))\n",
    "print('#Test set size : {}'.format(len(X_test)))\n",
    "\n",
    "okt = Okt()\n",
    "\n",
    "def twit_tokenizer(text):  # 명사, 동사, 형용사를 사용\n",
    "    target_tags = ['Noun', 'Verb', 'Adjective']\n",
    "    result = []\n",
    "    for word, tag in okt.pos(text, norm=True, stem=True):\n",
    "        if tag in target_tags:\n",
    "            result.append(word)\n",
    "    return result\n",
    "\n",
    "# 명사, 동사, 형용사를 이용해 tfidf 생성\n",
    "tfidf = TfidfVectorizer(tokenizer=twit_tokenizer,\n",
    "                        max_features=2000, min_df=5, max_df=0.5)\n",
    "X_train_tfidf = tfidf.fit_transform(X_train)\n",
    "X_test_tfidf = tfidf.transform(X_test)\n",
    "\n",
    "# 다중회귀분석 실시\n",
    "lr = LinearRegression()\n",
    "lr.fit(X_train_tfidf, y_train)\n",
    "print('#Regression Train set R2 score : {}'.format(lr.score(X_train_tfidf, y_train)))\n",
    "print('#Regression Test set R2 score : {}'.format(lr.score(X_test_tfidf, y_test)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#Train precision score : 0.8931011681728014\n",
      "#Test precision score : 0.867862969004894\n",
      "\n",
      "#Train recall score : 0.968682763566818\n",
      "#Test recall score : 0.9456096693921081\n",
      "\n",
      "#Train f1 score : 0.9293577981651375\n",
      "#Test f1 score : 0.905069751616196\n",
      "\n",
      "#Train accuracy score : 0.8884361133749887\n",
      "#Test accuracy score : 0.8484519282998371\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score\n",
    "\n",
    "y_train_senti = (y_train > 5)\n",
    "y_test_senti = (y_test > 5)\n",
    "\n",
    "y_train_predict = (lr.predict(X_train_tfidf) > 5)\n",
    "y_test_predict = (lr.predict(X_test_tfidf) > 5)\n",
    "\n",
    "# 정밀도\n",
    "train_precision_score = precision_score(y_train_senti, y_train_predict)\n",
    "test_precision_score = precision_score(y_test_senti, y_test_predict)\n",
    "\n",
    "# 재현율\n",
    "train_recall_score = recall_score(y_train_senti, y_train_predict)\n",
    "test_recall_score = recall_score(y_test_senti, y_test_predict)\n",
    "\n",
    "# f1-score\n",
    "train_f1_score = f1_score(y_train_senti, y_train_predict)\n",
    "test_f1_score = f1_score(y_test_senti, y_test_predict)\n",
    "\n",
    "# 정확도\n",
    "train_accuracy_score = accuracy_score(y_train_senti, y_train_predict)\n",
    "test_accuracy_score = accuracy_score(y_test_senti, y_test_predict)\n",
    "\n",
    "print('#Train precision score : {}'.format(train_precision_score))\n",
    "print('#Test precision score : {}\\n'.format(test_precision_score))\n",
    "\n",
    "print('#Train recall score : {}'.format(train_recall_score))\n",
    "print('#Test recall score : {}\\n'.format(test_recall_score))\n",
    "\n",
    "print('#Train f1 score : {}'.format(train_f1_score))\n",
    "print('#Test f1 score : {}\\n'.format(test_f1_score))\n",
    "\n",
    "print('#Train accuracy score : {}'.format(train_accuracy_score))\n",
    "print('#Test accuracy score : {}'.format(test_accuracy_score))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#Accuracy for train set: 0.878\n",
      "#Precision for train set: 0.878\n",
      "#Recall for train set: 0.973\n",
      "#F1 for train set: 0.923\n",
      "#Accuracy for test set: 0.855\n",
      "#Precision for test set: 0.866\n",
      "#Recall for test set: 0.958\n",
      "#F1 for test set: 0.910\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression #sklearn이 제공하는 logistic regression을 사용\n",
    "\n",
    "#count vector에 대해 regression을 해서 NB와 비교\n",
    "LR_clf = LogisticRegression() #분류기 선언\n",
    "LR_clf.fit(X_train_tfidf, y_train_senti) # train data를 이용하여 분류기를 학습\n",
    "\n",
    "y_train_predict = LR_clf.predict(X_train_tfidf)\n",
    "y_test_predict = LR_clf.predict(X_test_tfidf)\n",
    "\n",
    "print('#Accuracy for train set: {:.3f}'.format(accuracy_score(y_train_senti, y_train_predict)))\n",
    "print('#Precision for train set: {:.3f}'.format(precision_score(y_train_senti, y_train_predict)))\n",
    "print('#Recall for train set: {:.3f}'.format(recall_score(y_train_senti, y_train_predict)))\n",
    "print('#F1 for train set: {:.3f}'.format(f1_score(y_train_senti, y_train_predict)))\n",
    "\n",
    "print('#Accuracy for test set: {:.3f}'.format(accuracy_score(y_test_senti, y_test_predict)))\n",
    "print('#Precision for test set: {:.3f}'.format(precision_score(y_test_senti, y_test_predict)))\n",
    "print('#Recall for test set: {:.3f}'.format(recall_score(y_test_senti, y_test_predict)))\n",
    "print('#F1 for test set: {:.3f}'.format(f1_score(y_test_senti, y_test_predict)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "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": "3.9.13"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}