基于方面-词性感知的方面级情感分析

doi:10.19678/j.issn.1000-3428.0067204

摘要/Abstract

摘要：

方面级情感分析是自然语言处理的研究热点之一，其任务目的是预测句子中给定方面的情感极性。目前已有研究大多忽略了方面词和特定词性单词在过滤情感极性相关上下文语义信息和理解上下文语法信息中的作用。为此，提出一种基于方面-词性感知的图卷积网络ASP_POSGCN。采用双向长短期记忆网络建模上下文和词性信息，经由门控机制筛选方面词相关上下文语义信息，再使用词性信息隐藏层状态进一步过滤；同时设计方面-词性感知矩阵算法，根据不同词性单词对方面词情感极性的贡献重构句子原始依存关系以获取重构依存句法图，将原始依存句法图和重构依存句法图应用于双通道图卷积网络和多图感知机制；最后，使用过滤后的上下文语义信息与双通道图卷积网络的输出计算注意力得到最终分类表示。实验结果表明，该模型在Twitter、Laptop14、Restaurant14和Restaurant16 4个公开数据集上的准确率分别为74.57%、79.15%、83.84%、91.23%，F1值分别为72.59%、75.76%、77.00%、77.11%，与传统方面级情感分析基准模型相比均有提升，有助于方面级的情感极性分类。

关键词: 方面级情感分析, 图卷积网络, 门控机制, 词性信息, 多图感知机制

Abstract:

Aspect-Based Sentiment Analysis(ABSA) is a research focal point in natural language processing, whose task is to predict the emotional polarity of a given aspect in a sentence. Currently, most studies have overlooked the role of aspect words and specific Part-Of-Speech(POS) words in filtering contextual semantic information related to emotional polarity and understanding contextual grammatical information. Accordingly, a Graph Convolutional Network(GCN) based on aspect POS perception, ASP_POSGCN, is proposed. It adopts a bidirectional Long Short-Term Memory(LSTM) network to model context and POS information, filtering contextual semantic information related to aspect words through a gate mechanism, and further filtering using the state of the POS information hiding layer. Simultaneously, it designs an aspect POS perception matrix algorithm to reconstruct the original dependency relationship of sentences based on the contribution of different POS words to the emotional polarity of aspect words, to obtain the reconstructed dependency syntax graph. The original and reconstructed dependency syntax graphs are applied to the dual-channel GCN and multi-graph perception mechanism. Finally, it uses the filtered contextual semantic information and the output of the dual channel GCN, and attention is calculated to obtain the final classification representation. The experimental results demonstrate that the accuracy of the model on four public datasets, Twitter, Laptop14, Restaurant14 and Restaurant16, is 74.57%, 79.15%, 83.84%, and 91.23%, and the F1 values are 72.59%, 75.76%, 77.00%, and 77.11%, respectively. Compared with the traditional ABSA benchmark model, it has improved and is beneficial to sentiment polarity classification in aspects.

Key words: Aspect-Based Sentiment Analysis(ABSA), Graph Convolution Network(GCN), gate mechanism, Part-Of-Speech(POS) information, multi-graph perception mechanism

夏卫欢, 廖列法, 张守信, 张燕琴. 基于方面-词性感知的方面级情感分析[J]. 计算机工程, 2024, 50(3): 68-77.

Weihuan XIA, Liefa LIAO, Shouxin ZHANG, Yanqin ZHANG. Aspect-Based Sentiment Analysis Based on Aspect-Part-Of-Speech Perception[J]. Computer Engineering, 2024, 50(3): 68-77.

https://www.ecice06.com/CN/Y2024/V50/I3/68

图/表 18

图1 ASP_POSGCN模型结构

Fig.1 Structure of ASP_POSGCN model

图2 长短期记忆网络结构

Fig.2 Structure of long short-term network

图3 双向长短期记忆网络结构

Fig.3 Structure of bi-directional long short-term network

图4 原始依存句法树

Fig.4 Primitive dependency syntax tree

图5 依存句法树重构

Fig.5 Reconstruction of dependency syntax tree

图6 消融模型与准确率的关系

Fig.6 The relationship between ablation models and accuracy

图7 消融模型与F1值的关系

Fig.7 The relationship between ablation models and F1 value

图8 特征相关系数

$ {\alpha } $

与准确率的关系

Fig.8 The relationship between characteristic correlation coefficient

$ {\alpha } $

and accuracy

图9 特征相关系数

$ {\alpha } $

与F1值的关系

Fig.9 The relationship between characteristic correlation coefficient

$ {\alpha } $

and F1 value

图10 词性嵌入维度与准确率的关系

Fig.10 The relationship between the embedding dimension of part-of-speech and accuracy

图11 词性嵌入维度与F1值的关系

Fig.11 The relationship between the embedding dimension of part-of-speech and F1 value

图12 GCN层数与准确率的关系

Fig.12 The relationship between GCN layers and accuracy

图13 GCN层数与F1值的关系

Fig.13 The relationship between GCN layers and F1 value

参考文献 24

1	ZHANG S Q, XU X Y, PANG Y W, et al. Multi-layer attention based CNN for target-dependent sentiment classification. Neural Processing Letters, 2020, 51 (3): 2089- 2103. doi: 10.1007/s11063-019-10017-9
2	XUE W, LI T. Aspect based sentiment analysis with gated convolutional networks[EB/OL]. [2023-02-13]. https://arxiv.org/abs/1805.07043v1.
3	TANG D Y, QIN B, FENG X C, et al. Effective LSTMs for target-dependent sentiment classification[EB/OL]. [2023-02-13]. https://arxiv.org/pdf/1512.01100.pdf.
4	WANG Y Q, HUANG M L, ZHAO L, et al. Attention-based LSTM for aspect-level sentiment classification[EB/OL]. [2023-02-13]. https://aclanthology.org/D16-1058.pdf.
5	MA D H, LI S J, ZHANG X D, et al. Interactive attention networks for aspect-level sentiment classification[EB/OL]. [2023-02-13]. https://arxiv.org/pdf/1709.00893.pdf.
6	CHEN P, SUN Z Q, BING L D, et al. Recurrent attention network on memory for aspect sentiment analysis[C]//Proceedings of Conference on Empirical Methods in Natural Language Processing. Stroudsburg, USA: Association for Computational Linguistics, 2017: 452-461.
7	CHEN H, YAN Y, DU S. Research on aspect-level sentiment analysis of user reviews. Journal of Frontiers of Computer Science & Technology, 2021, 15 (3): 478.
8	LIU M Z, ZHOU F Y, CHEN K, et al. Co-attention networks based on aspect and context for aspect-level sentiment analysis. Knowledge-Based Systems, 2021, 217, 106810. doi: 10.1016/j.knosys.2021.106810
9	ZHANG C, LI Q C, SONG D W. Aspect-based sentiment classification with aspect-specific graph convolutional networks[EB/OL]. [2023-02-13]. https://arxiv.org/abs/1909.03477v1.
10	XIAO L W, HU X H, CHEN Y N, et al. Targeted sentiment classification based on attentional encoding and graph convolutional networks. Applied Sciences, 2020, 10 (3): 957. doi: 10.3390/app10030957
11	王汝言, 陶中原, 赵荣剑, 等. 多交互图卷积网络用于方面情感分析. 电子与信息学报, 2022, 44 (3): 1111- 1118. URL
	WANG R Y, TAO Z Y, ZHAO R J, et al. Multi-interaction graph convolutional networks for aspect-level sentiment analysis. Journal of Electronics & Information Technology, 2022, 44 (3): 1111- 1118. URL
12	唐晓波, 刘广超. 细粒度情感分析综述. 图书情报工作, 2017, 61 (5): 132- 140. URL
	TANG X B, LIU G C. Research review on fine-grained sentiment analysis. Library and Information Service, 2017, 61 (5): 132- 140. URL
13	GU T Q, ZHAO H, HE Z, et al. Integrating external knowledge into aspect-based sentiment analysis using graph neural network. Knowledge-Based Systems, 2023, 259, 110025. doi: 10.1016/j.knosys.2022.110025
14	KHONG W H, SOON L K, GOH H N, et al. Leveraging part-of-speech tagging for sentiment analysis in short texts and regular texts[C]//Proceedings of Joint International Semantic Technology Conference. Berlin, Germany: Springer, 2018: 182-197.
15	SHUANG K, GU M Y, LI R, et al. Interactive POS-aware network for aspect-level sentiment classification. Neurocomputing, 2021, 420, 181- 196. doi: 10.1016/j.neucom.2020.08.013
16	KIM D, KIM J, JEONG Y S. Graph convolutional networks with POS gate for aspect-based sentiment analysis. Applied Sciences, 2022, 12 (19): 10134. doi: 10.3390/app121910134
17	TIAN Y H, CHEN G M, SONG Y. Aspect-based sentiment analysis with type-aware graph convolutional networks and layer ensemble[C]//Proceedings of Conference of the North American Chapter of the Association for Computational Linguistics. Stroudsburg, USA: Association for Computational Linguistics, 2021: 2910-2922.
18	LIANG B, SU H, GUI L, et al. Aspect-based sentiment analysis via affective knowledge enhanced graph convolutional networks. Knowledge-Based Systems, 2022, 235, 107643. doi: 10.1016/j.knosys.2021.107643
19	ZHANG M, QIAN T. Convolution over hierarchical syntactic and lexical graphs for aspect level sentiment analysis[C]//Proceedings of Conference on Empirical Methods in Natural Language Processing. Stroudsburg, USA: Association for Computational Linguistics, 2020: 3540-3549.
20	JIANG L, YU M, ZHOU M, et al. Target-dependent Twitter sentiment classification[C]//Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies. New York, USA: ACM Press, 2011: 151-160.
21	KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[C]//Proceedings of the 5th International Conference on Learning Representations. Stroudsburg, USA: Association for Computational Linguistics, 2017: 1-10.
22	TANG D Y, QIN B, LIU T. Aspect level sentiment classification with deep memory network[EB/OL]. [2023-02-13]. https://arxiv.org/pdf/1605.08900.pdf.
23	HUANG B, OU Y, CARLEY K M. Aspect level sentiment classification with attention-over-attention neural networks[C]//Proceedings of the 11th International Conference on Social, Cultural, and Behavioral Modeling. Washington D. C., USA: IEEE Press, 2018: 197-206.
24	LIU J, LIU P Y, ZHU Z F, et al. Graph convolutional networks with bidirectional attention for aspect-based sentiment classification. Applied Sciences, 2021, 11 (4): 1528. doi: 10.3390/app11041528

[1]	代巍, 王丰羽, 冀常鹏. 基于情感增强与双图卷积网络的方面级情感分析[J]. 计算机工程, 2024, 50(5): 120-127.
[2]	杨春霞, 吴亚雷, 闫晗, 黄昱锟. 融合双图卷积与门控线性单元的方面级情感分析模型[J]. 计算机工程, 2024, 50(4): 141-149.
[3]	莫少聪, 陈庆锋, 谢泽, 刘春雨, 邱俊铼. 基于动态图注意力与标签传播的实体对齐[J]. 计算机工程, 2024, 50(4): 150-159.
[4]	刘威, 马磊, 李凯, 李蓉. 基于多粒度字形增强的中文医学命名实体识别[J]. 计算机工程, 2024, 50(2): 337-344.
[5]	马建红, 龚天, 姚爽. 基于证据句与图卷积网络的文档级关系抽取[J]. 计算机工程, 2023, 49(8): 104-110.
[6]	郭艳霞, 金勇, 唐宏, 彭金枝. 基于动态卷积与残差门控的多模态情感识别[J]. 计算机工程, 2023, 49(7): 94-101.
[7]	代祖华, 刘园园, 狄世龙. 语义增强的图神经网络方面级文本情感分析[J]. 计算机工程, 2023, 49(6): 71-80.
[8]	陈昱瑾, 王晶, 武志昊, 赵耀帅, 林友芳. 基于图卷积网络融合群组关系的社会化推荐方法[J]. 计算机工程, 2023, 49(5): 112-121.
[9]	陈文轩, 曾碧, 郭植星. 融合多特征与语义图卷积网络的摔倒检测方法[J]. 计算机工程, 2023, 49(5): 277-285,294.
[10]	陈柏霖, 王天极, 任丽娜, 黄瑞章. 融合ELECTRA和文本局部信息的中文语法错误检测方法[J]. 计算机工程, 2023, 49(3): 304-311.
[11]	袁立宁, 胡皓, 刘钊. 基于多通道图卷积自编码器的图表示学习[J]. 计算机工程, 2023, 49(2): 150-160,174.
[12]	杨海洋, 张兴鹏. 融合多通道图卷积网络的方面级情感分析模型[J]. 计算机工程, 2023, 49(11): 61-69.
[13]	刘宽, 奚小冰, 周明东. 基于自适应多尺度图卷积网络的骨架动作识别[J]. 计算机工程, 2023, 49(10): 264-271.
[14]	张文豪, 廖列法, 王茹霞. 融合句法树多信息学习方面级情感分析[J]. 计算机工程, 2023, 49(10): 72-79.
[15]	孔博, 韩虎, 陈景景, 白雪, 邓飞. 基于虚拟依存关系与知识增强的方面级情感分析[J]. 计算机工程, 2023, 49(10): 53-63.

选择文件类型/文献管理软件名称

选择包含的内容