融合词法句法信息的方面级情感分析模型  

Aspect-level sentiment analysis model incorporating lexical and syntactic information

作　　者：衡红军[1] 杨鼎诚 HENG Hong-jun;YANG Ding-cheng(College of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300

出　　处：《计算机工程与设计》2024年第3期837-844,共8页Computer Engineering and Design

基　　金：国家自然科学基金联合基金项目(U1333109)。

摘　　要：为解决现有方面级情感分析方法缺乏句法约束和词义信息的问题,将句法依存树和知识图谱融合起来对句子编码,提出一种词法句法相结合的图神经网络模型。利用图神经网络分别提取句法依存树中的句法信息和知识图谱中的词法信息,经过位置编码模块和掩码加权模块捕捉重要性更高的单词;将两种特征进行结合获得融合句法词法信息的文本表示,进行情感分类。在3个公开数据集上的实验结果验证了该模型的有效性。To solve the problem that the existing aspect-level sentiment analysis methods lack syntactic constraints and lexical information,the syntactic dependency tree and knowledge graph were innovatively integrated to encode sentences,and a graph neural network model combining lexical syntax was proposed.The syntactic information in the syntactic dependency tree and the lexical information in the knowledge graph were extracted using the graph neural network,and the words with higher importance were captured using the position encoding module and the mask weighting module.The two features were combined to obtain a fusion syntactic lexical text representation of information for sentiment classification.Experimental results on three public datasets validate the effectiveness of the model.

关 键 词：方面级情感分析 句法约束 词义信息 句法依存树 知识图谱 关系图注意力网络 图卷积网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]