Edited at

たぶん1分くらいでできる形態素解析とtfidf(テストコードつき)

More than 5 years have passed since last update.


下準備

pip install nltk

pip install mecab-python


下のコードを貼りつけて実行してみてください

TF-IDFを出力するための関数がtfidf

形態素解析するための関数がextract_words

下の方にあるimport unittest以下の長ったらしいヤツはテスト

#!/usr/bin/env python

#-*- encoding: utf-8 -*-
import nltk
import MeCab
import urllib2
from urllib2 import HTTPError
from itertools import chain

def tfidf(doc,docs):
"""対象の文書と全文の形態素解析した単語リストを指定すると対象の文書のTF-IDFを返す"""
tokens = list(chain.from_iterable(docs)) #flatten
A = nltk.TextCollection(docs)
token_types = set(tokens)
return [{"word":token_type,"tfidf":A.tf_idf(token_type, doc)} for token_type in token_types]

def extract_words(text):
"""テキストを与えると名詞のリストにして返す"""
text = text.encode("utf-8") if isinstance(text,unicode) else text
mecab = MeCab.Tagger("")
node = mecab.parseToNode(text)
words = []
while node:
fs = node.feature.split(",")
if (node.surface is not None) and node.surface != "" and fs[0] in [u'名詞']:
words.append(node.surface)
node = node.next
return words

import unittest

class MachineLearningTest(unittest.TestCase):
def test_extract_words(self):
"""形態素解析のテスト"""
text = "textを形態素解析して、名詞のリストを返す"
keywords = extract_words(text)
self.assertEqual(keywords, ["text","形態素","解析","名詞","リスト"])
def test_tfidf(self):
"""tfidfのテスト"""
urls = ["http://qiita.com/puriketu99/items/"+str(i) for i in range(1,10)]
def url2words(url):
try:
html = urllib2.urlopen(url).read()
except HTTPError:
html = ""
plain_text = nltk.clean_html(html).replace('\n','')
words = extract_words(plain_text)
return words
docs = [url2words(url) for url in urls]
tfidfs_fizzbuzz = tfidf(docs[0],docs)
tfidfs_fizzbuzz.sort(cmp=lambda x,y:cmp(x["tfidf"],y["tfidf"]),reverse=True)
result = [e for i,e in enumerate(tfidfs_fizzbuzz) if len(e["word"]) > 2 and i < 30]
self.assertEqual(result[7]["word"],"yaotti")#Qiita側がデザイン変えるとテスト失敗するかも
print result
#[{'tfidf': 0.08270135278254376, 'word': 'quot'},
# {'tfidf': 0.02819364299404901, 'word': 'FizzBuzz'},
# {'tfidf': 0.02067533819563594, 'word': 'fizzbuzz'},
# {'tfidf': 0.02067533819563594, 'word': 'Buzz'},
# {'tfidf': 0.016916185796429405, 'word': 'Fizz'},
# {'tfidf': 0.016726267030018446, 'word': 'end'},
# {'tfidf': 0.015036609596826138, 'word': 'map'},
# {'tfidf': 0.015036609596826138, 'word': 'yaotti'},
# {'tfidf': 0.011277457197619604, 'word': 'def'}]

if __name__ == '__main__':
unittest.main()


参考

TF-IDFの計算

http://everydayprog.blogspot.jp/2011/12/tf-idf.html