word2vec全部弹幕比较句子相似度情感五分类

#!/usr/bin/python
# -*- coding: UTF-8 -*-
import sys
import codecs   #可以以特定编码打开文件
import jieba
import jieba.posseg as pseg
reload(sys)               #zzh说这种方法不好，不要再用了！！！  可是真的很好用啊 QAQ
sys.setdefaultencoding('utf-8')
import gensim

#model = gensim.models.Word2Vec.load("22620491.model")
model = gensim.models.KeyedVectors.load_word2vec_format('news_12g_baidubaike_20g_novel_90g_embedding_64.bin', binary=True)
word_vec = model.wv
del model     #把模型给word_vec，所以Model删掉。
'''
print word_vec[u'难过']
'''


f = codecs.open("qinggancidanmu.txt",'r','utf-8')    #codecs包指定TXT打开方式
lines = f.readlines()
doc = open('fenlei.txt', 'w')




for line in lines: #每一行弹幕
    if lines.index(line) % 500 ==0:   #显示跑到多少条数据
        print lines.index(line)

    words=line.split(" ")
    words.pop(0)
    u = []
    for word in words:
            if word != "\r\n":    #去掉换行符，linux只用\n换行。win下用\r\n表示换行。反正\n不行就\r\n试试！
                #print type(word)
                u.append(word)    #word_vec输入必须要unicode才行。



    le = [u'乐']
    ai = [u'哀']
    nu = [u'怒']
    jing = [u'惊']
    wu = [u'恶']
    try:
        l,a,n,j,w=word_vec.n_similarity(u, le),word_vec.n_similarity(u, ai),word_vec.n_similarity(u, nu),word_vec.n_similarity(u, jing),word_vec.n_similarity(u, wu)
        list=[l,a,n,j,w]
        #print list
        doc.write(line.split(" ",1)[0]+" "+str(list.index(max(list)))+" "+line.split(" ",1)[1]+'\n')      # index记得 变成 str啊 ！！改了好半天!



    except:
        doc.write(line.split(" ",1)[0]+" "+"-1"+" "+line.split(" ",1)[1]+'\n')
        continue



print("end")
f.close()
doc.close()