语义和句法双增强的交互式方面级情感分析  

Semantic and syntactic dual-enhanced interactive attention networks for aspect-level sentiment analysis

作　　者：王法玉[1] 邱意雯 陈洪涛[1] WANG Fa-yu;QIU Yi-wen;CHEN Hong-tao(Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology,Tianjin University of Technology,Tianjin 300384,China)

机构地区：[1]天津理工大学智能计算及软件新技术天津市重点实验室,天津300384

出　　处：《计算机工程与设计》2024年第12期3719-3725,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61571328);天津理工大学教学研究与改革基金项目(YB22-12)。

摘　　要：针对目前研究存在的语义和句法信息提取不充分以及忽略二者交互的问题,提出一种利用交互注意力机制融合语义和句法信息的方面级情感分析模型。将方面间依赖关系与局部语义融合,获得综合的全局语义信息,将全局与局部信息进行交互得到更深层的语义信息;利用改进的图卷积网络增强模型提取上下文句法信息的能力;使用多头交互注意力完成方面词与上下文之间以及增强语义和句法之间的交互。为验证模型的有效性,在Laptop14、Restaurat14和Twitter基准数据集上进行实验,实验结果表明,所提模型取得的性能优于比较方法。Aiming at the problems of insufficient extraction of semantic and syntactic information and neglect of the interaction between the two,an aspect-level sentiment analysis model using the interactive attention mechanism to integrate semantic and syntactic information was proposed.The inter-aspect dependency was fused with local semantics to obtain comprehensive global semantic information,and global and local information was interacted to obtain deeper semantic information.The improved graph convolutional network was used to enhance the model’s ability to extract contextual syntactic information.Multi-head interaction attention was used to complete interactions between aspect words and context,as well as the enhanced semantics and syntax.To verify the effectiveness of the model,experiments were carried out on the Laptop14,Restaurat14 and Twitter benchmark datasets.Experimental results show that the proposed model achieves better performance than the comparison method.

关 键 词：方面级情感分析 深度学习 自然语言处理 交互式注意力机制 图卷积网络 语义增强 句法增强 句法分析 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]