结合句法增强的多通道方面级情感分析模型  

Syntactic enhanced multi-channel aspect-based emotion analysis model

作　　者：牛利月 郑秋生[1,2] 张龙 王鹏 NIU Liyue;ZHENG Qiusheng;ZHANG Long;WANG Peng(Research Institute of Frontier Information Technology,Zhongyuan University of Technology,Zhengzhou 450007,China;Henan Key Laboratory on Public Opinion Intelligent Analysis,Zhengzhou 450000,China)

机构地区：[1]中原工学院前沿信息技术研究院,郑州450007 [2]河南省网络舆情智能检测与分析重点实验室,郑州450000

出　　处：《智能计算机与应用》2022年第8期48-53,共6页Intelligent Computer and Applications

基　　金：河南省网络舆情与智能分析重点实验室与河南省教育厅重点科研项目(22B520054);国家自然科学基金(61702547)。

摘　　要：方面级情感分析(ABSA)作为情感分析中的一项精细任务,旨在分析给定方面在特定语境中的情感极性。目前广泛使用的情感方法,大部分基于深度神经网络提取语义信息或者句法信息,在准确建模方面,词与观点词之间的关系存在不足。为解决上述问题,本文提出了结合句法增强的多通道方面级情感分析模型。该模型借助依存句法树,对句子中特定方面及其观点词进行建模,同时采用单词共现的方法构建单词共现图;分别使用图卷积神经网络提取图特征,同时利用Bi-GRU提取句子语义特征,最终将特征融合进行情感分类。经在3个经典数据集上验证,证明了该模型的有效性。As a fine-grained task in emotion analysis,aspect level sentiment analysis(ABSA)aims to analyze the emotional polarity of a given aspect in a specific context.Most of the currently widely used methods are based on deep neural networks(DNNs)to extract semantic or syntactic information,however there are shortcomings in modeling the relation between the aspects and its opinion words accurately.In order to solve the above problems,a multi-channel aspect-level sentiment analysis model combined with syntactic enhancement is proposed.The model uses the dependency syntax tree to model specific aspects of the sentence and its opinion words.At the same time,the word co-occurrence method is used to construct the word co-occurrence graph,and the graph convolutional neural networks(GCN)is used to extract the graph features.At the same time,Bi-GRU,which is good at capturing the full-text features of the text context,is used to extract the semantic features of the sentence.Finally,the features are fused for sentiment classification.Experimental results on three classical data sets show the effectiveness of the proposed model.

关 键 词：方面级情感分析 依存句法树 单词共现 图卷积神经网络 Bi-GRU 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]