【摘要】 针对目前微博平台的文本情感分类模型大多是对句子的词性、表情符号等进行情感分析而不了解用户本身的情感倾向且存在语义理解不足的问题,提出一种利用Word2Vec结合深度学习的方法对微博用户进行情感分类。使用Word2Vec中的Skip-Gram模型结合负采样对语料训练词向量,然后利用双向长短期记忆网络(Bi LSTM)-ATT模型自动学习词向量中的情感信息,捕捉文本数据中最具代表性的特征,最后经过Soft Max层对微博用户的情感倾向进行分类。在NLPCC2013数据集上进行测试,同时做了5组对比试验。结果表明:所提出的模型AVP达到0.814,AVF1值达到0.831,且在词向量维度取150时效果最好。更多还原

【Abstract】 In view of the current emotional classification models for text on weibo platform,most of them conduct emotional analysis on the part of speech or emoji of a setence,etc. without understanding users ’ emotional tendency and lack of semantic understanding. Proposing a method of emotional classification for weibo users by using Word2 Vec combined with deep learning. Specifically,skip-gram model in Word2 Vec and negative sampling are used to train word vectors. Then bidirectional long short-term memory( Bi LSTM)-attention mechanism( ATT)model is used to automatically learn the emotional information in word vectors,capture the most representative features in text data,and finally classify the emotional tendencies of Weibo users through softmax layer. Test on NLPCC2013 data set,at the same time,made five group of contrast test. The results show that the new model AVP reaches 0. 814,AVF1 reaches 0. 831,and the effect is best when the dimension of word vector is 150.更多还原