《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (9): 2752-2760.DOI: 10.11772/j.issn.1001-9081.2025080979

• 人工智能 • 上一篇    

基于异构增强与多源知识融合的方面级情感分析

王伊璠1, 韩虎2(), 李栋2, 范雅婷2, 李琳1   

  1. 1.兰州交通大学 数理学院,兰州 730070
    2.兰州交通大学 电子与信息工程学院,兰州 730070
  • 收稿日期:2025-08-25 修回日期:2025-10-17 接受日期:2025-10-20 发布日期:2026-09-16 出版日期:2026-09-10
  • 通讯作者: 韩虎
  • 作者简介:王伊璠(1999—),女,山西运城人,硕士研究生,主要研究方向:自然语言处理、方面级情感分析
    韩虎(1977—),男,甘肃兰州人,教授,博士,主要研究方向:数据挖掘、自然语言处理
    李栋(2000—),男,安徽亳州人,硕士研究生,主要研究方向:自然语言处理、多模态情感分析
    范雅婷(1999—),女,甘肃兰州人,硕士,主要研究方向:自然语言处理
    李琳(1998—),女,甘肃酒泉人,硕士,主要研究方向:情感词典构建。
  • 基金资助:
    国家自然科学基金资助项目(62166024);甘肃省重点研发计划-工业类项目(25YFGA046)

Heterogeneous enhancement and multi-source knowledge fusion for aspect-based sentiment analysis

Yifan WANG1, Hu HAN2(), Dong LI2, Yating FAN2, Lin LI1   

  1. 1.School of Mathematics and Physics,Lanzhou Jiaotong University,Lanzhou Gansu 730070,China
    2.School of Electronic and Information Engineering,Lanzhou Jiaotong University,Lanzhou Gansu 730070,China
  • Received:2025-08-25 Revised:2025-10-17 Accepted:2025-10-20 Online:2026-09-16 Published:2026-09-10
  • Contact: Hu HAN
  • About author:WANG Yifan, born in 1999, M. S. candidate. Her research interests include natural language processing, aspect-based sentiment analysis.
    HAN Hu, born in 1977, Ph. D., professor. His research interests include data mining, natural language processing.
    LI Dong, born in 2000, M. S. candidate. His research interests include natural language processing, multimodal sentiment analysis.
    FAN Yating, born in 1999, M. S. Her research interests include natural language processing.
    LI Lin, born in 1998, M. S. Her research interests include emotional dictionary construction.
  • Supported by:
    National Natural Science Foundation of China(62166024);Key Research and Development Program of Gansu Province - Industrial Category(25YFGA046)

摘要:

方面级情感分析(ABSA)是自然语言处理中的重要研究任务,现有方法普遍采用单一类型节点建模句子中方面词与意见词的语义关联,难以有效区分不同类型节点对情感表达的差异性贡献。此外,现有的模型通常利用句法依赖解析器构建句法树进行语义分析,忽视了单词本身所蕴含的细粒度语义信息及文本中隐含的情感强度修饰关系。针对上述问题,提出一种基于异构增强与多源知识融合的图卷积网络(HMS-GCN)模型。首先,构建基于差异化约束的异构图网络,利用异构图中的多类型节点及其相应边关系,多维度捕捉句子的语义情感信息,并通过双通道注意力机制关注不同类型节点及节点内部的重要程度;其次,利用义原知识对初始文本特征进行语义增强,并在句法邻接矩阵的基础上融合层级短语的结构信息,从而提升模型对句法依存关系的理解能力;最后,融合异构图网络信息和知识增强特征进行情感分类预测。在公开数据集Twitter、Lap14、Rest15和Rest16上的实验结果表明,相较于ISSK-GCN(Integrating Syntax and Sentiment Knowledge Graph Convolutional Network)等模型,所提模型的准确率取得了最优结果,验证了差异化节点建模与知识增强策略的有效性。

关键词: 方面级情感分析, 异构图, 双通道注意力机制, 图卷积网络, 义原知识

Abstract:

Aspect-Based Sentiment Analysis (ABSA) is an important task in natural language processing. The existing methods generally adopt single-type nodes to model semantic associations between aspects and opinion terms, failing to distinguish the contributions of different node types to sentiment representation effectively. Moreover, current models usually construct syntactic trees using dependency parsers for semantic analysis, ignoring fine-grained semantic information of the word itself and implicit sentiment intensity modification relations in the text. To address the above issues, a Heterogeneous enhancement and Multi-Source knowledge fusion Graph Convolutional Network (HMS-GCN) model was proposed. First, a heterogeneous graph with differentiated constraints was constructed to capture semantic and sentiment information in sentences multi-dimensionally by using the relations of multi-type nodes and their corresponding edges in the heterogeneous graph, and importance of different types of nodes and their internal components were focused on by utilizing a dual-channel attention mechanism. Second, sememe knowledge was adopted to enhance the semantic representation of initial textual features, and hierarchical phrase structural information was integrated on the basis of the syntactic adjacency matrix, so as to improve the model’s syntactic dependency comprehension. Finally, the heterogeneous graph network information and enhanced knowledge features were fused for sentiment classification prediction. Experimental results on public datasets (Twitter, Lap14, Rest15, Rest16) show that compared to models such as ISSK-GCN (Integrating Syntax and Sentiment Knowledge Graph Convolutional Network), the proposed model achieves the best results, validating the effectiveness of the differentiated node modeling and knowledge enhancement strategies.

Key words: Aspect-Based Sentiment Analysis (ABSA), heterogeneous graph, dual-channel attention mechanism, Graph Convolutional Network (GCN), sememe knowledge

中图分类号: