期刊文献+

基于注意力双层LSTM的长文本情感分类方法 被引量:4

Long Text Emotion Classification Method based on the Attention Double-layer LSTM
下载PDF
导出
摘要 针对由于长文本评论篇幅长,正负情感特征离散分布且每个句子的的情感语义贡献度不同,导致难以准确判断全文情感倾向的问题,提出一种基于注意力双层LSTM的长文本情感倾向性分析方法。该方法首先利用LSTM学习句子级情感向量表示;然后采用双向LSTM对文档中所有句子的情感语义及句子间的语义关系进行编码,并基于注意力机制对具有不同情感语义贡献度的句子进行权值分配;最后,加权句子级情感向量表示得到长文本的文档级情感向量表示,经过Softmax层得到长文本情感倾向。在Yelp2015和IMDb电影评论语料上实验,结果表明该方法能达到较好的分类效果,进一步提升了情感分类的正确率。 Due to the long length and complex structure of the long text, the discrete distribution of positive and negative sentiment features, and the different affective semantic contributions of each sentence,it is difficult to accurately judge the emotional tendency of the full text. A long text sentiment analysis method based on attention mechanism and bilayer LSTM is proposed. Firstly, LSTM is used to learn sentence level emotion vector representation. Then the emotional semantics of all sentences in the document and the semantic relations between sentences are encoded by bidirectional LSTM, and the weights of sentences with different emotional semantic contributions are assigned based on the attention mechanism. Finally, the weighted sentence level emotion vector represents the document level emotion vector representation of the long text, and gets the long text emotion tendency through the Softmax layer. Experiments on Yelp2015 and IMDb film review corpus show that this method can achieve better classification effect and further improve the accuracy of emotion classification.
作者 毛焱颖 MAO Yanying(Chongqing College of Electronic Engineering, Chongqing 401331, China)
出处 《重庆电子工程职业学院学报》 2019年第2期118-125,共8页 Journal of Chongqing College of Electronic Engineering
关键词 长短时记忆网络 注意力机制 评论文本 情感分类 long and short time memory network attention mechanism comment text sentiment classification
  • 相关文献

参考文献1

二级参考文献6

共引文献29

同被引文献50

引证文献4

二级引证文献16

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部