nltk-比较中文文档类似度-完整实例

摘要: 对比于《nltk-比较英文文档类似度-完整实例》html

nltk同时也能处理中文的场景,只要作以下改动:python

  1. 使用中文分词器(如我选用告终巴分词)app

  2. 对中文字符作编码处理,使用unicode编码方式dom

  3. python的源码编码统一声明为 gbkide

  4. 使用支持中文的语料库this

 

 

 

代码以下,须要jieba的支持编码

 

#!/usr/bin/env pythonspa

#-*-coding=gbk-*-日志

 

"""code

     原始数据,用于创建模型

"""

#缩水版的courses,实际数据的格式应该为 课程名\t课程简介\t课程详情,并已去除html等干扰因素

courses = [           

            u'Writing II: Rhetorical Composing',

            u'Genetics and Society: A Course for Educators',

            u'General Game Playing',

            u'Genes and the Human Condition (From Behavior to Biotechnology)',

            u'A Brief History of Humankind',

            u'New Models of Business in Society',

            u'Analyse Numrique pour Ingnieurs',

            u'Evolution: A Course for Educators',

            u'Coding the Matrix: Linear Algebra through Computer Science Applications',

            u'The Dynamic Earth: A Course for Educators',

            u'Tiny Wings\tYou have always dreamed of flying - but your wings are tiny. Luckily the world is full of beautiful hills. Use the hills as jumps - slide down, flap your wings and fly! At least for a moment - until this annoying gravity brings you back down to earth. But the next hill is waiting for you already. Watch out for the night and fly as fast as you can. ',

            u'Angry Birds Free',

            u'没有\它很类似',

            u'没有\t它很类似',

            u'没有\t他很类似',

            u'没有\t他不很类似',

            u'没有',

            u'能够没有',

            u'也没有',

            u'有没有也无论',

            u'Angry Birds Stella',

            u'Flappy Wings - FREE\tFly into freedom!A parody of the #1 smash hit game!',

            u'没有一个',

            u'没有一个2',

           ]

 

#只是为了最后的查看方便

#实际的 courses_name = [course.split('\t')[0] for course in courses]

courses_name = courses

 

 

"""

    预处理(easy_install nltk)

"""

def pre_process_cn(courses, low_freq_filter = True):

    """

     简化的 中文+英文 预处理

        1.去掉停用词

        2.去掉标点符号

        3.处理为词干

        4.去掉低频词

 

    """

    import nltk

    import jieba.analyse

    from nltk.tokenize import word_tokenize

   

    texts_tokenized = []

    for document in courses:

        texts_tokenized_tmp = []

        for word in word_tokenize(document):

            texts_tokenized_tmp += jieba.analyse.extract_tags(word,10)

        texts_tokenized.append(texts_tokenized_tmp)   

   

    texts_filtered_stopwords = texts_tokenized

 

    #去除标点符号

    english_punctuations = [',', '.', ':', ';', '?', '(', ')', '[', ']', '&', '!', '*', '@', '#', '$', '%']

    texts_filtered = [[word for word in document if not word in english_punctuations] for document in texts_filtered_stopwords]

 

    #词干化

    from nltk.stem.lancaster import LancasterStemmer

    st = LancasterStemmer()

    texts_stemmed = [[st.stem(word) for word in docment] for docment in texts_filtered]

   

    #去除太低频词

    if low_freq_filter:

        all_stems = sum(texts_stemmed, [])

        stems_once = set(stem for stem in set(all_stems) if all_stems.count(stem) == 1)

        texts = [[stem for stem in text if stem not in stems_once] for text in texts_stemmed]

    else:

        texts = texts_stemmed

    return texts

 

lib_texts = pre_process_cn(courses)

 

 

 

"""

    引入gensim,正式开始处理(easy_install gensim)

"""

 

def train_by_lsi(lib_texts):

    """

        经过LSI模型的训练

    """

    from gensim import corpora, models, similarities

 

    #为了能看到过程日志

    #import logging

    #logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)

 

    dictionary = corpora.Dictionary(lib_texts)

    corpus = [dictionary.doc2bow(text) for text in lib_texts]     #doc2bow(): 将collection words 转为词袋,用两元组(word_id, word_frequency)表示

    tfidf = models.TfidfModel(corpus)

    corpus_tfidf = tfidf[corpus]

 

    #拍脑壳的:训练topic数量为10的LSI模型

    lsi = models.LsiModel(corpus_tfidf, id2word=dictionary, num_topics=10)

    index = similarities.MatrixSimilarity(lsi[corpus])     # index 是 gensim.similarities.docsim.MatrixSimilarity 实例

   

    return (index, dictionary, lsi)

 

   

#库创建完成 -- 这部分可能数据很大,能够预先处理好,存储起来

(index,dictionary,lsi) = train_by_lsi(lib_texts)

   

   

#要处理的对象登场

target_courses = [u'没有']

target_text = pre_process_cn(target_courses, low_freq_filter=False)

 

 

"""

对具体对象类似度匹配

"""

 

#选择一个基准数据

ml_course = target_text[0]

 

#词袋处理

ml_bow = dictionary.doc2bow(ml_course)  

 

#在上面选择的模型数据 lsi 中,计算其余数据与其的类似度

ml_lsi = lsi[ml_bow]     #ml_lsi 形式如 (topic_id, topic_value)

sims = index[ml_lsi]     #sims 是最终结果了, index[xxx] 调用内置方法 __getitem__() 来计算ml_lsi

 

#排序,为输出方便

sort_sims = sorted(enumerate(sims), key=lambda item: -item[1])

 

#查看结果

print sort_sims[0:10]   #看下前10个最类似的,第一个是基准数据自身

print courses_name[sort_sims[1][0]]   #看下实际最类似的数据叫什么

print courses_name[sort_sims[2][0]]   #看下实际最类似的数据叫什么

print courses_name[sort_sims[3][0]]   #看下实际最类似的数据叫什么

相关文章
相关标签/搜索