联合con-GRU与ATGAT模型的情感分析三元组方法  

A triple method for emotional analysis using con⁃GRU and ATGAT models

作　　者：毕晓杰 李卫疆[1,2] BI Xiaojie;LI Weijiang(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Key Laboratory of Artificial Intelligence,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500 [2]昆明理工大学云南省人工智能重点实验室,云南昆明650500

出　　处：《现代电子技术》2024年第8期149-154,共6页Modern Electronics Technique

摘　　要：情感分析三元组任务是情感分析任务的研究热点,其目的在于将方面词、情感词与情感极性组成三元组。图神经网络是提取句子特征的有效手段,但其无法关注节点与节点之间的关系,并且注意力权重赋予不合理。为此,提出一种联合双向对抗GRU与基于句法注意力机制的GAT模型。利用依存句法树的句子向量和ATGAT模型提取句子情感词,并将句子用syn-str更新向量表示;再融合句法结构的向量,通过双向对抗GRU提取特征进行情感判断。在三个公开英文数据集中开展实验,结果表明,所提模型相对其他基线模型具有更优异的性能。消融实验和对比实验也说明,所提出的各个网络模型组件比其他组件能更有效地将句法信息与原句向量融合。The emotional analysis triplet task is a hot research topic in emotional analysis tasks,with the aim of combining aspect words,emotional words,and emotional polarity into triplets.Graph neural networks are an effective means of extracting sentence features,but they cannot pay attention to the relationships between nodes during the process,and the allocation of attention weights is unreasonable.A GAT model joint bidirectional adversarial GRU and based on syntactic attention mechanism is proposed.The sentence vectors from dependency syntax trees and the ATGAT model are used to extract sentence sentiment words,and the sentences are represented by syn-str update vectors.The experimental results on three publicly available English datasets show that the proposed model has better performance compared with the other baseline models.The ablation and comparative experiments also demonstrate that the proposed network model components can more effectively fuse syntactic information with the original sentence vector than other components.

关 键 词：情感分析 三元组 双向对抗GRU GAT模型 句法注意力机制 依存句法树 特征提取 

分 类 号：TN919-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]