#!/usr/bin/python
# -*- coding: UTF-8 -*-
import sys
import codecs   #可以以特定编码打开文件
import numpy as np

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




for line in lines: #每一行弹幕
    if lines.index(line) % 100 ==0:   #显示跑到多少条数据
        print lines.index(line)
    list=[]
    for i in range(0,64):
        list.append(0)
    list=np.array(list)          #转化为array便于加减乘除
    words=line.split("  ",1)[1].split(" ")
    count=0
    for word in words:
        if word != "\r\n":
            print(word)
        try:
            print(word_vec[word])
            count=count+1
            print(type(word_vec[word]))
            print(type(list))
            list=word_vec[word]+list    #向量对应值相加
            print(list)
            print(count)
        except:
            continue

    print(count)
    if count !=0:
        list=list/count

    print(list)

    list=list.tolist()    # 变回List容易增加项 和 SVM分类
    list.append(int(float(line.split("  ",1)[0].split(" ",5)[2])))
    list.append(int(float(line.split("  ", 1)[0].split(" ", 5)[3])))
    list.append(int(float(line.split("  ", 1)[0].split(" ", 5)[4])))
    list.append(int(float(line.split("  ", 1)[0].split(" ", 5)[5])))
    print(list)
    doc.write(line.split(" ",2)[0]+" "+str(list)+"\r\n")




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