微博作为当今热门的社交平台,其中蕴含着许多具有强烈主观性的用户评论文本.为挖掘微博评论文本中潜在的信息,针对传统的情感分析模型中存在的语义缺失以及过度依赖人工标注等问题,提出一种基于LSTM+Word2vec的深度学习情感分析模型.采用Word2vec中的连续词袋模型(continuous bag of words,CBOW),利用语境的上下文结构及语义关系将每个词语映射为向量空间,增强词向量之间的稠密度;采用长短时记忆神经网络模型实现对文本上下文序列的线性抓取,最后输出分类预测的结果.实验结果的准确率可达95.9％,通过对照实验得到情感词典、RNN、SVM三种模型的准确率分别为52.3％、92.7％、85.7％,对比发现基于LSTM+Word2vec的深度学习情感分析模型的准确率更高,具有一定的鲁棒性和泛化性,对用户个性化推送和网络舆情监控具有重要意义.
Sentiment analysis of Weibo comments based on LSTM+Word2vec
Weibo is a popular social platform,contains many subjective user comments.In order to explore the potential information in the comment text of weibo,a deep learning sentiment analysis model based on LSTM+Word2vec is proposed to solve the problems of semantic loss and excessive dependence on manual annotation in the traditional sentiment analysis model.The CBOW(continuous bag of words)model in Word2vec is used to map words into vector space by using the context structure and semantic relationship of context,so as to enhance the density between word vectors.LSTM is used to realize the linear capture of the text context sequence,and finally yield the result of classification prediction.As a control experiment,the accuracy of the three models of sentiment dictionary,RNN and SVM is 52.3％,92.7％and 85.7％respectively.It is found that the accuracy of the deep learning sentiment analysis model based on LSTM+Word2vec is higher,which has certain robustness and generalization.It is of great significance for user personalized push and network public opinion monitoring.
万方数据
维普
