面向目标情感分析的双重图注意力网络模型

doi:10.19665/j.issn1001-2400.2023.01.016

摘要/Abstract

摘要：

目标情感分析旨在分析评论文本中不同目标所对应的情感倾向。当前,基于图神经网络的方法使用依存句法树来融入依存句法关系,一方面,此类方法大多忽略了依存关系缺乏区分度的事实;另一方面,未考虑依存句法树提供的依存关系存在目标与情感词关系缺失的问题。为此,提出双重图注意力网络模型。该模型首先使用双向长短期记忆网络得到具有语义信息的词节点表示,然后根据依存句法树在词节点表示上构建句法图注意力网络,实现依存句法关系重要程度的区分,更有效地建立目标与情感词之间的关系,进而得到更准确的目标情感特征表示;同时根据句子的无向完全图构建全局图注意力网络来挖掘目标与情感词缺失的关系,进一步提升模型的性能。实验结果表明,与现有模型对比,双重图注意力网络模型在不同数据集上的准确率与宏平均F1值均取得了更好结果。

关键词: 自然语言处理, 目标情感分析, 依存关系, 图注意力网络, 注意力机制

Abstract:

Target sentiment analysis aims to analyze the sentiment tendency corresponding to different targets in the review text.At present,graph neural network based methods use the dependency syntactic tree to incorporate dependency syntactic relations.On the one hand,these methods mostly ignore the fact that dependency relations lack distinction.On the other hand,without considering the dependency relations provided by the dependency syntactic tree,there is a lack of relations between target and sentiment words.Therefore,a dual graph attention network(DGAT) model is proposed.First,the model uses a bidirectional long short-term memory network to obtain word node representation with semantic information,and then constructs a syntactic graph attention network based on the word node representation according to the dependency syntactic tree,so as to distinguish the importance of dependency syntactic relations,more effectively establish the relation between target and sentiment words,and obtain a more accurate representation of target sentiment features.At the same time,according to the undirected complete graph of sentences,a global graph attention network is used to mine lacking relations between target and sentiment words,so as to further improve the performance of the model.Experimental results show that compared with existing models,the DGAT model has a better accuracy and macro-average F1 value on different datasets.

Key words: natural language processing, target sentiment analysis, dependency relations, graph attention network, attention mechanism

中图分类号:

TP391

崔少国,陈思奇,杜兴. 面向目标情感分析的双重图注意力网络模型[J]. 西安电子科技大学学报, 2023, 50(1): 137-148.

CUI Shaoguo,CHEN Siqi,DU Xing. Dual graph attention networks model for target sentiment analysis[J]. Journal of Xidian University, 2023, 50(1): 137-148.

图/表 20

图1

图2

图3

图4

图5

图6

表1

表2

表3

图7

图8

图9

图10

图11

图12

图13

图14

图15

图16

图17

参考文献 25

[1]	ZHOU J, HUANG J X, CHEN Q, et al. Deep Learning for Aspect-Level Sentiment Classification:Survey,Vision,and Challenges[J]. IEEE Access, 2019, 7:78454-78483. doi: 10.1109/Access.6287639
[2]	曹卫东, 李嘉琪, 王怀超. 采用注意力门控卷积网络模型的目标情感分析[J]. 西安电子科技大学学报, 2019, 46(6):30-36.
	CAO Weidong, LI Jiaqi, WANG Huaichao. Analysis of Targeted Sentiment by the Attention Gated Convolutional Network Model[J]. Journal of Xidian University, 2019, 46(6):30-36.
[3]	ZHU X, ZHU L, GUO J, et al. GL-GCN:Global and Local Dependency Guided Graph Convolutional Networks for Aspect-Based Sentiment Classification[J]. Expert Systems with Applications, 2021, 186:115712. doi: 10.1016/j.eswa.2021.115712
[4]	XU Q, ZHU L, DAI T, et al. Aspect-Based Sentiment Classification with Multi-Attention Network[J]. Neurocomputing, 2020, 388:135-143. doi: 10.1016/j.neucom.2020.01.024
[5]	WANG K, SHEN W, YANG Y, et al. Relational Graph Attention Network for Aspect-Based Sentiment Analysis[C]// Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.Online:ACL, 2020:3229-3238.
[6]	TANG D, QIN B, FENG X, et al. Effective LSTMs for Target-Dependent Sentiment Classification[C]// Proceedings of COLING 2016,the 26th International Conference on Computational Linguistics:Technical Papers.Osaka:ACL, 2016:3298-3307.
[7]	GRAVES A, JAITLY N, MOHAMED A. Hybrid Speech Recognition with Deep Bidirectional LSTM[C]// 2013 IEEE Workshop on Automatic Speech Recognition and Understanding.Olomouc:ASRU, 2013:273-278.
[8]	WANG Y, HUANG M, ZHU X, et al. Attention-Based LSTM for Aspect-Level Sentiment Classification[C]// Proceedings of the 2016Conference on Empirical Methods in Natural Language Processing.Austin:EMNLP, 2016:606-615.
[9]	MA D, LI S, ZHANG X, et al. Interactive Attention Networks for Aspect-level Sentiment Classification[C]// Proceedings of the 26th International Joint Conference on Artificial Intelligence.Melbourne:IJCAI, 2017:4068-4074.
[10]	CHEN P, SUN Z, BING L, et al. Recurrent Attention Network on Memory for Aspect Sentiment Analysis[C]// Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing.Copenhagen:EMNLP, 2017:452-461.
[11]	GU S, ZHANG L, HOU Y, et al. A Position-Aware Bidirectional Attention Network for Aspect-Level Sentiment Analysis[C]// Proceedings of the 27th International Conference on Computational Linguistics.Santa Fe, New Mexico: COLING, 2018:774-784.
[12]	FAN F, FENG Y, ZHAO D. Multi-grained Attention Network for Aspect-level Sentiment Classification[C]// Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing.Brussels:EMNLP, 2018:3433-3442.
[13]	HUANG B, OU Y, CARLEY K M. Aspect Level Sentiment Classification with Attention-Over-Attention Neural Networks[C]// International Conference on Social Computing,Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation. Washington DC: SBP-BRiMS, 2018:197-206.
[14]	ZHANG C, LI Q, SONG D. Aspect-based Sentiment Classification with Aspect-Specific Graph Convolutional Networks[C]// Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing. Hong Kong: EMNLP-IJCNLP, 2019:4568-4578.
[15]	SUN K, ZHANG R, MENSAH S, et al. Aspect-Level Sentiment Analysis via Convolution over Dependency Tree[C]// Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing. Hong Kong: EMNLP-IJCNLP, 2019:5679-5688.
[16]	HUANG B, CARLEY K M. Syntax-Aware Aspect Level Sentiment Classification with Graph Attention Networks[C]// Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing. Hong Kong: EMNLP-IJCNLP, 2019:5469-5477.
[17]	王光, 李鸿宇, 邱云飞, 等. 基于图卷积记忆网络的方面级情感分类[J]. 中文信息学报, 2021, 35(8):98-106.
	WANG Guang, LI Hongyu, QIU Yunfei, et al. Aspect-Based Sentiment Classification via Memory Graph Convolutional Network[J]. Journal of Chinese Information Processing, 2021, 35(8):98-106.
[18]	王汝言, 陶中原, 赵容剑, 等. 多交互图卷积网络用于方面情感分析[J]. 电子与信息学报, 2022, 44(3):1111-1118.
	WANG Ruyan, TAO Zhongyuan, ZHAO Rongjian, et al. Multi-Interaction Graph Convolutional Networks for Aspect-Level Sentiment Analysis[J]. Journal of Electronics & Information Technology, 2022, 44(3):1111-1118.
[19]	PENNINGTON J, SOCHER R, MANNING C D. Glove:Global Vectors for Word Representation[C]// Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing.Doha:EMNLP, 2014:1532-1543.
[20]	李源, 崔玉爽, 王伟. 一种基于字词双通道网络的文本情感分析方法[J]. 西安电子科技大学学报, 2021, 48(6):179-186.
	LI Yuan, CUI Yushuang, WANG Wei. Method for the Analysis of Text Sentiment Based on the Word Dual-Channel Network[J]. Journal of Xidian University, 2021, 48(6):179-186.
[21]	VELIČKOVIĆ P, CUCURULL G, CASANOVA A, et al. Graph Attention Networks[J]. The 6th International Conference on Learning Representations.Vancouver:ICLR, 2018:1-12.
[22]	ZHOU J, HUANG J X, HU Q V, et al. Sk-Gcn:Modeling Syntax and Knowledge via Graph Convolutional Network for Aspect-Level Sentiment Classification[J]. Knowledge-Based Systems, 2020, 205:106292. doi: 10.1016/j.knosys.2020.106292
[23]	蒋浩泉, 张儒清, 郭嘉丰, 等. 图卷积网络与自注意机制在文本分类任务上的对比分析[J]. 中文信息学报, 2021, 35(12):84-93.
	JIANG Haoquan, ZHANG Ruqing, GUO Jiafeng, et al. A Comparative Study of Graph Convolutional Networks and Self-Attention Mechanism on Text Classification[J]. Journal of Chinese Information Processing, 2021, 35(12):84-93.
[24]	DONG L, WEI F, TAN C, et al. Adaptive Recursive Neural Network for Target-Dependent Twitter Sentiment Classification[C]// Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics(Volume 2:Short Papers).Baltimore:ACL, 2014:49-54.
[25]	PONTIKI M, GALANIS D, PAVLOPOULOS J, et al. Semeval-2014 Task 4:Aspect Based Sentiment Analysis[C]// The 8th International Workshop on Semantic Evaluation.Dublin:SemEval, 2014:27-35.

数据集	积极		中性		消极
数据集	训练集	测试集	训练集	测试集	训练集	测试集
Twitter	1 561	173	3 127	346	1 560	173
Restaurant	2 164	728	637	196	807	196
Laptop	994	341	464	169	870	128

超参数	数值或名称
词嵌入维度	300
LSTM隐藏层输出	300
最大迭代次数epoch	100
批次大小batch size	32
L2正则化因子λ	0.000 01
LeakyRelu负半轴斜率	0.01
优化器	Adam
学习率	0.001

类别	模型	Twitter		Laptop		Restaurant
类别	模型	Acc	MF1	Acc	MF1	Acc	MF1
基线	LSTM	68.64	66.84	67.71	60.69	76.96	63.35
注意力模型	ATAE-LSTM^[8]			68.70		77.20
	IAN^[9]			72.10		78.60
	RAM^[10]	69.36	67.30	74.49	71.35	80.23	70.80
	PBAN^[11]			74.12		81.16
	MGAN^[12]	72.54	70.81	75.39	72.47	81.25	71.94
	AOA-LSTM^[13]			74.50		81.20
GCN模型	ASGCN^[14]	72.15	70.40	75.55	71.05	80.77	72.02
	MemGCN^[17]	72.19	70.29	74.38	70.30	81.55	72.87
	MIGCN^[18]	73.31	72.12	76.59	72.44	82.32	74.31
文中模型	DGAT	73.99	72.53	76.49	72.75	82.68	75.53