摘要
在互联网平台上,用户可以针对电影、新闻等发表自己的观点、表达自己的情感,为其他用户提供消费该商品的参考意见,也帮助产品经理制定有效的产品消费策略。目前,针对中文情感倾向分类、深度学习的方法取得了一定的成就,尤其是长短期记忆神经网络(LSTM)。该网络是一个时序模型,可以很好地理解评论语义抓住评论中蕴含的情感倾向,但是它存在词向量构建阶段无法突出情感词的情感信息,以及无法针对不同场景进行文本情感倾向分析的问题。为此,提出LSTM前融合情感倾向分类模型。新模型利用情感词的情感标签修正情感词向量,解决了情感词向量无法突出情感信息的问题,并且将电影的简介作为一个输入特征融合到最终句子的特征向量中,实现针对具体的电影新闻场景评论情感倾向分类。实验结果表明,新模型相对于基本的LSTM模型取得了更好的效果,亦表明该模型能更加精确地抓取评论的情感信息。
Users can express their own opinions and emotions on the Internet platform for movies,news,etc.,provide other users with reference opinions on consumption of the product,and help product strategists to formulate effective product consumption strategies.At present,although for the classification of Chinese affective tendencies,deep learning method has made achievements,especially Long Short-Term Memory Model(LSTM).The network is a time series model,which can well understand the comment semantics and grasp the emotional tendency contained in the comment.However,it has the problems that the emotional information of emotional words cannot be highlighted in the construction stage of word vector and the emotional tendency analysis of text cannot be carried out for different scenes.For this,a LSTM pre-fusion emotional tendency classification model is proposed in this paper,which uses emotional labels of emotional words to modify the emotional words vector,solves the problem that emotional words vector can not highlight emotional information,and integrates the brief introduction of the movie as an input feature into the final sentence feature vector to achieve the emotional tendency classification of specific movie news scene comments.The experimental results show that the novel model is better than the basic LSTM model,and can capture the emotion information of comments more accurately.
作者
王铃
陶宏才
WANG Ling;TAO Hongcai(School of Information Science&Technology,Southwest Jiaotong University,Chengdu 611756,China)
出处
《成都信息工程大学学报》
2020年第2期139-145,共7页
Journal of Chengdu University of Information Technology
基金
国家自然科学基金资助项目(61806170)。
关键词
情感倾向分类
LSTM
情感词向量
前融合模型
classification of emotional tendency
LSTM
emotional words vector
pre-fusion model
