多源知识融合的方面级情感分析模型

韩虎; 郝俊; 张千锟; 赵启涛

doi:10.13700/j.bh.1001-5965.2022.0728

多源知识融合的方面级情感分析模型

doi: 10.13700/j.bh.1001-5965.2022.0728

兰州交通大学电子与信息工程学院，兰州 730070

基金项目: 国家自然科学基金(62166024)

详细信息

通讯作者:
E-mail：hanhu_lzjtu@mail.lzjtu.cn

中图分类号: TP391
计量
- 文章访问数: 343
- HTML全文浏览量: 139
- PDF下载量: 43
- 被引次数: 0
出版历程
- 收稿日期: 2022-08-17
- 录用日期: 2023-01-14
- 网络出版日期: 2023-03-23
- 整期出版日期: 2024-09-27

Multi-source knowledge fusion model for aspect-based sentiment analysis

School of Electronic and Information Engineering，Lanzhou Jiaotong University，Lanzhou 730070，China

Funds: National Natural Science Foundation of China (62166024)

More Information

Corresponding author: E-mail：hanhu_lzjtu@mail.lzjtu.cn

摘要

摘要:
方面级情感分析(ABSA)是一项细粒度情感分析任务，其目的是针对评论语句中出现的特定方面给出对应的情感极性。现有的基于深度学习的ABSA方法大多侧重于评论语句语义和句法的挖掘，往往忽略了评论语句可能涉及的概念知识和情感程度信息。针对此问题，提出一种融合多源知识的神经网络模型，通过句法依赖揭示句子的结构框架、词共现捕捉单词之间的语义联系、情感网络和概念图谱的嵌入为模型提供情感和背景知识，共同实现评论语句上下文与评价方面的增强表示，并通过双交互注意力模式实现评论语句上下文与评价方面的协调优化。通过在4个公开数据集上的实验验证，该模型在ABSA任务中，准确率分别达到了75.00%、77.90%、81.55%、90.10%，与基准模型相比均有所提高。研究成果不仅验证了多源知识融合在ABSA任务中的有效性，也为未来的研究提供了新的思路和方法。
- 方面级情感分析 /
- 图卷积网络 /
- 多源融合 /
- 知识图谱 /
- 交互注意力机制
Abstract:
Aspect-based sentiment analysis (ABSA) is a fine-grained sentiment analysis task that aims to give the corresponding sentiment polarity for specific aspects that appear in review statements. Most existing aspect-based sentiment analysis methods relying on deep learning focus on mining the semantics and syntax of review statements, often ignoring the conceptual knowledge and sentiment degree information that may be involved in the review statements. To address this problem, a neural network model incorporating multi-source knowledge. It reveals the structural framework of sentences through syntactic dependencies, captures semantic connections between words through word co-occurrence, and embeds emotional networks and concept graphs to provide emotional and background knowledge for the model, and coordinated optimization of the contextual and evaluative aspects of review statements was realized through a dual-interaction attention model. Experimental results on four public datasets show that the model achieves better performance than existing models, with accuracy reaching 75.00%, 77.90%, 81.55%, and 90.10%, respectively, all of which were improved compared to the benchmark model. This achievement not only verifies the effectiveness of multi-source knowledge fusion in ABSA tasks, but also provides new ideas and methods for future research.
- aspect-based sentiment analysis /
- graph convolutional network /
- multi-source fusion /
- knowledge graph /
- interactive attention mechanisms

HTML全文

图 1 依存句法树^[5]

Figure 1. Dependency syntax tree^[5]

下载: 全尺寸图片幻灯片

图 2 层次图^[8]

Figure 2. Hierarchical graph^[8]

下载: 全尺寸图片幻灯片

图 3 MSKFSA 模型网络整体框架

Figure 3. Overall framework of MSKFSA model

下载: 全尺寸图片幻灯片

图 4 GCN层数与准确率的关系

Figure 4. Relationship between number of GCN layers and accuracy

下载: 全尺寸图片幻灯片

图 5 GCN层数与Marco-F₁的关系

Figure 5. Relationship between number of GCN layers and Marco-F₁

下载: 全尺寸图片幻灯片

表 1 ABSA任务数据集

Table 1. ABSA task datasets

数据集	训练集个数			测试集个数
数据集	积极	中性	消极	积极	中性	消极
Twitter	1561	3127	1560	173	346	173
Lap14	994	464	870	341	169	128
Rest15	912	36	256	326	34	182
Rest16	1240	69	439	469	30	117

下载: 导出CSV

表 2 不同模型结果对比

Table 2. Results of different models %

类别	模型	准确率				Marco-F₁
类别	模型	Twitter	Lap14	Rest15	Rest16	Twitter	Lap14	Rest15	Rest16
LSTM模型	LSTM^[4]	69.56	69.28	77.37	86.80	67.70	63.09	55.17	63.88
LSTM模型	RAM^[29]	69.36	74.49			67.30	71.35
交互模型	IAN^[13]	72.50	72.05	78.54	84.74	70.81	67.38	52.65	55.21
	MGAN^[30]	72.54	75.27			70.81	70.81
	AOA^[31]	72.30	76.62	78.17	87.50	70.20	67.52	57.02	66.21
GCN模型	ASGCN^[5]	72.15	75.55	79.89	88.99	70.40	71.05	61.89	67.48
	BiGCN^[8]	74.16	74.59	81.16	88.96	73.35	71.84	64.79	70.84
	GL-GCN^[32]	73.26	76.91	80.81	88.47	71.26	72.76	64.99	69.64
	MIGCN^[34]	73.31	76.59	80.81	89.50	72.12	72.44	64.21	71.97
	KEATGCN^[33]	74.57	76.65	80.07	89.45	73.34	73.21	65.90	74.81
本文模型	MSKFSA	75.00	77.90	81.55	90.10	74.03	74.14	67.47	71.99

下载: 导出CSV

表 3 消融实验结果

Table 3. Results of ablation experiments %

模型	准确率				Marco-F₁
模型	Twitter	Lap14	Rest15	Rest16	Twitter	Lap14	Rest15	Rest16
MSKFSA	75.00	77.90	81.55	90.10	74.03	74.14	67.47	71.99
w/o lan	73.99	72.26	79.89	87.18	72.37	67.28	62.81	68.95
w/o em	74.13	72.10	78.78	87.01	72.29	67.16	64.83	64.78
w/o co	74.42	70.85	78.60	86.69	72.91	64.71	60.06	67.65
w/o in	74.57	76.65	80.63	88.96	73.03	72.95	66.05	67.51

下载: 导出CSV

参考文献(34)

[1]	LIU B, ZHANG L. A survey of opinion mining and sentiment analysis[M]//AGGARWAL C C, ZHAI C X. Mining text data. Berlin: Springer, 2012: 415-463.
[2]	唐晓波, 刘广超. 细粒度情感分析研究综述[J]. 图书情报工作, 2017, 61(5): 132-140. TANG X B, LIU G C. Research review on fine-grained sentiment analysis[J]. Library and Information Service, 2017, 61(5): 132-140(in Chinese).
[3]	KIM Y. Convolutional neural networks for sentence classification[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2014: 1746-1751.
[4]	TANG D Y, QIN B, FENG X C. Effective LSTMs for target-dependent sentiment classification[C]//Proceedings of the 26th International Conference on Computational Linguistics. Stroudsburg: ACL, 2016: 3298-3307.
[5]	ZHANG C, LI Q C, SONG D W. Aspect-based sentiment classification with aspect-specific graph convolutional networks[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2019: 4567-4577.
[6]	WANG K, SHEN W Z, YANG Y 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. Stroudsburg: ACL, 2020: 3229-3238.
[7]	谢珺, 王雨竹, 陈波, 等. 基于双指导注意力网络的属性情感分析模型[J]. 计算机研究与发展, 2022, 59(12): 2831-2843. doi: 10.7544/issn1000-1239.20210708 XIE J, WANG Y Z, CHEN B, et al. Aspect-based sentiment analysis model with bi-guide attention network[J]. Journal of Computer Research and Development, 2022, 59(12): 2831-2843(in Chinese). doi: 10.7544/issn1000-1239.20210708
[8]	ZHANG M, QIAN T Y. Convolution over hierarchical syntactic and lexical graphs for aspect level sentiment analysis[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2020: 3540-3549.
[9]	MA Y K, PENG H Y, CAMBRIA E. Targeted aspect-based sentiment analysis via embedding commonsense knowledge into an attentive LSTM[C]//Proceedings of the 32nd Conference on Artificial Intelligence. Palo Alto: AAAI, 2018: 5876-5883.
[10]	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
[11]	ZHAO A P, YU Y. Knowledge-enabled BERT for aspect-based sentiment analysis[J]. Knowledge-Based Systems, 2021, 227: 107220. doi: 10.1016/j.knosys.2021.107220
[12]	WANG Y Q, HUANG M L, ZHU X Y, et al. Attention-based LSTM for aspect-level sentiment classification[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2016: 606-615.
[13]	MA D H, LI S J, ZHANG X D, et al. Interactive attention networks for aspect-level sentiment classification[C]//Proceedings of the 26th International Joint Conference on Artificial Intelligence. New York: ACM, 2017: 4068-4074.
[14]	PANG S G, XUE Y, YAN Z H, et al. Dynamic and multi-channel graph convolutional networks for aspect-based sentiment analysis[C]//Proceedings of the Findings of the Association for Computational Linguistics. Stroudsburg: ACL, 2021: 2627-2636.
[15]	LI R F, CHEN H, FENG F X, et al. Dual graph convolutional networks for aspect-based sentiment analysis[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2021: 6319-6329.
[16]	ZHONG Q H, DING L, LIU J H, et al. Knowledge graph augmented network towards multiview representation learning for aspect-based sentiment analysis[EB/OL]. (2022-01-13)[2022-08-01].
[17]	CHEN J D, HU Y Z, LIU J P, et al. Deep short text classification with knowledge powered attention[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 6252-6259. doi: 10.1609/aaai.v33i01.33016252
[18]	BIAN X M, FENG C, AHMAD A, et al. Targeted sentiment classification with knowledge powered attention network[C]//Prroceedings of the IEEE 31st International Conference on Tools with Artificial Intelligence. Piscataway: IEEE Press, 2019: 1073-1080.
[19]	CAMBRIA E, SPEER R, HAVASI C, et al. SenticNet: A publicly available semantic resource for opinion mining[C]//Proceedings of the AAAI Fall Symposium. Palo Alto: AAAI, 2010: 14-18.
[20]	CAMBRIA E, LI Y, XING F Z, et al. SenticNet 6: Ensemble application of symbolic and subsymbolic AI for sentiment analysis[C]//Proceedings of the 29th ACM International Conference on Information & Knowledge Management. New York: ACM, 2020: 105-114.
[21]	XING F Z, PALLUCCHINI F, CAMBRIA E. Cognitive-inspired domain adaptation of sentiment lexicons[J]. Information Processing and Management, 2019, 56(3): 554-564. doi: 10.1016/j.ipm.2018.11.002
[22]	LIANG B, SU H, GUI L, et al. Aspect-based sentiment analysis via affective knowledge enhanced graph convolutional networks[J]. Knowledge-Based Systems, 2022, 235: 107643. doi: 10.1016/j.knosys.2021.107643
[23]	PENNINGTON J, SOCHER R, MANNING C. GloVe: Global vectors for word representation[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2014: 1532-1543.
[24]	JI L, WANG Y J, SHI B T, et al. Microsoft concept graph: Mining semantic concepts for short text understanding[J]. Data Intelligence, 2019, 1(3): 238-270. doi: 10.1162/dint_a_00013
[25]	DONG L, WEI F R, TAN C Q, 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. Stroudsburg: ACL, 2014: 49-54.
[26]	PONTIKI M, GALANIS D, PAVLOPOULOS J, et al. SemEval-2014 task 4: Aspect based sentiment analysis[C]//Proceedings of the 8th International Workshop on Semantic Evaluation. Stroudsburg: ACL, 2014: 27-35.
[27]	PONTIKI M, GALANIS D, PAPAGEORGIOU H, et al. SemEval-2015 task 12: Aspect based sentiment analysis[C]//Proceedings of the 9th International Workshop on Semantic Evaluation. Stroudsburg: ACL, 2015: 486-495.
[28]	PONTIKI M, GALANIS D, PAPAGEORGIOU H, et al. SemEval-2016 task 5: Aspect based sentiment analysis[C]//Proceedings of the 10th International Workshop on Semantic Evaluation. Stroudsburg: ACL, 2016: 19-30.
[29]	CHEN P, SUN Z Q, BING L D, et al. Recurrent attention network on memory for aspect sentiment analysis[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2017: 452-461.
[30]	FAN F F, FENG Y S, ZHAO D Y. Multi-grained attention network for aspect-level sentiment classification[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2018: 3433-3442.
[31]	HUANG B X, OU Y L, CARLEY K M. Aspect level sentiment classification with attention-over-attention neural networks[M]// THOMSON R, DANCY C, HYDER A, et al. Social, cultural, and behavioral modeling. Berlin: Springer, 2018: 197-206.
[32]	ZHU X F, ZHU L, GUO J F, 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
[33]	韩虎, 郝俊, 张千锟, 等. 知识增强的交互注意力方面级情感分析模型[J]. 计算机科学与探索, 2023, 17(3): 709-718. doi: 10.3778/j.issn.1673-9418.2108082 HAN H, HAO J, ZHANG Q K, et al. Knowledge-enhanced interactive attention model for aspect-based sentiment analysis[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(3): 709-718(in Chinese). doi: 10.3778/j.issn.1673-9418.2108082
[34]	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. doi: 10.11999/JEIT210459 WANG R Y, TAO Z Y, ZHAO R J, et al. Multi-interaction graph convolutional networks for aspect-level sentiment analysis[J]. Journal of Electronics & Information Technology, 2022, 44(3): 1111-1118(in Chinese). doi: 10.11999/JEIT210459