tf(term frequency)词频,在文章中出现次数最多的词,然而文章中出现次数较多的词并不一定就是关键词,比如常见的对文章本身并没有多大意义的停用词。所以我们需要一个重要性调整系数来衡量一个词是不是常见词。该权重为idf(inverse document frequency)逆文档频率,它的大小与一个词的常见程度成反比。在我们得到词频(tf)和逆文档频率(idf)以后,将两个值相乘,即可得到一个词的tf-idf值,某个词对文章的重要性越高,其tf-idf值就越大,所以排在最前面的几个词就是文章的关键词。
tf-idf算法的优点是简单快速,结果比较符合实际情况,但是单纯以“词频”衡量一个词的重要性,不够全面,有时候重要的词可能出现的次数并不多,而且这种算法无法体现词的位置信息,出现位置靠前的词和出现位置靠后的词,都被视为同样重要,是不合理的。
tf-idf算法步骤:
(1)、计算词频:
词频 = 某个词在文章中出现的次数
考虑到文章有长短之分,考虑到不同文章之间的比较,将词频进行标准化
词频 = 某个词在文章中出现的次数/文章的总词数
词频 = 某个词在文章中出现的次数/该文出现次数最多的词出现的次数
(2)、计算逆文档频率
需要一个语料库(corpus)来模拟语言的使用环境。
逆文档频率 = log(语料库的文档总数/(包含该词的文档数 + 1))
(3)、计算tf-idf
tf-idf = 词频(tf)* 逆文档频率(idf)
详细代码如下:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
|
#!/usr/bin/env python #-*- coding:utf-8 -*- ''' 计算文档的tf-idf ''' import codecs import os import math import shutil #读取文本文件 def readtxt(path): with codecs. open (path, "r" ,encoding = "utf-8" ) as f: content = f.read().strip() return content #统计词频 def count_word(content): word_dic = {} words_list = content.split( "/" ) del_word = [ "\r\n" , "/s" , " " , "/n" ] for word in words_list: if word not in del_word: if word in word_dic: word_dic[word] = word_dic[word] + 1 else : word_dic[word] = 1 return word_dic #遍历文件夹 def funfolder(path): filesarray = [] for root,dirs,files in os.walk(path): for file in files: each_file = str (root + "//" + file ) filesarray.append(each_file) return filesarray #计算tf-idf def count_tfidf(word_dic,words_dic,files_array): word_idf = {} word_tfidf = {} num_files = len (files_array) for word in word_dic: for words in words_dic: if word in words: if word in word_idf: word_idf[word] = word_idf[word] + 1 else : word_idf[word] = 1 for key,value in word_dic.items(): if key ! = " " : word_tfidf[key] = value * math.log(num_files / (word_idf[key] + 1 )) #降序排序 values_list = sorted (word_tfidf.items(),key = lambda item:item[ 1 ],reverse = true) return values_list #新建文件夹 def buildfolder(path): if os.path.exists(path): shutil.rmtree(path) os.makedirs(path) print ( "成功创建文件夹!" ) #写入文件 def out_file(path,content_list): with codecs. open (path, "a" ,encoding = "utf-8" ) as f: for content in content_list: f.write( str (content[ 0 ]) + ":" + str (content[ 1 ]) + "\r\n" ) print ( "well done!" ) def main(): #遍历文件夹 folder_path = r "分词结果" files_array = funfolder(folder_path) #生成语料库 files_dic = [] for file_path in files_array: file = readtxt(file_path) word_dic = count_word( file ) files_dic.append(word_dic) #新建文件夹 new_folder = r "tfidf计算结果" buildfolder(new_folder) #计算tf-idf,并将结果存入txt i = 0 for file in files_dic: tf_idf = count_tfidf( file ,files_dic,files_array) files_path = files_array[i].split( "//" ) #print(files_path) outfile_name = files_path[ 1 ] #print(outfile_name) out_path = r "%s//%s_tfidf.txt" % (new_folder,outfile_name) out_file(out_path,tf_idf) i = i + 1 if __name__ = = '__main__' : main() |
以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持服务器之家。
原文链接:https://blog.csdn.net/lalalawxt/article/details/79499498