方面情感三元组抽取的提示增强学习网络

doi:10.3778/j.issn.1002-8331.2405-0073

摘要/Abstract

摘要： 方面情感三元组抽取在最新的研究中致力于使用边界驱动的表格填充范式，然而其在预训练中学习到的表格特征和下游设计的解码方式之间缺乏语义交互，导致表格特征需要经历频繁的语义修改以适应任务目标。为了解决这个问题，提出一个提示增强的学习网络（prompt-enhanced learning network for aspect sentiment triplet extraction，PENet）进行方面情感三元组抽取。针对下游边界驱动的解码任务构造了新颖的边界标签提示，在上游特征学习中设计了相匹配的多粒度注意力卷积模块和提示改进的变压器结构，分别稳定地学习改进后文本的语义特征及在不同情感极性下面向任务的学习提示表示，同时基于提示不断调整文本特征的学习方向，最终实现提示范式引导下的三元组抽取网络，解决了解码前特征反复修改的问题。实验结果表明，在四个公开数据集上，PENet的综合表现已经超越了现有基线模型，显著提升了三元组抽取性能。

关键词: 方面情感三元组抽取, 提示学习, 注意力卷积

Abstract: Aspect sentiment triplet extraction in recent research is dedicated to using a boundary-driven table-filling paradigm, however, it lacks semantic interactions between the table features learned in pre-training and the decoding methods designed downstream, resulting in the need for the table features to undergo frequent semantic modifications to adapt to the task goals. To address this problem, a prompt-enhanced learning network (PENet) is proposed for aspect sentiment triplet extraction. A novel boundary label prompt is constructed for the downstream boundary-driven decoding task, and compatible multi-granularity attention convolution module and prompt-enhanced transformer structure are designed in the upstream feature learning, respectively, to stably learn the semantic features of improved text and acquire prompt representations under different sentiment polarities in the downstream task. Meanwhile, the learning direction of the text features is continuously adjusted base on the prompts. And the triplet extraction network under the guidance of the prompting paradigm is achieved, which solves the problem of repeated modification of the features before decoding. The experimental results show that the comprehensive performance of PENet has outperformed the existing baseline models on the four public datasets, significantly improving the triplet extraction performance.

Key words: aspect sentiment triplet extraction, prompt learning, attention convolution

李丹, 夏鸿斌, 刘渊. 方面情感三元组抽取的提示增强学习网络[J]. 计算机工程与应用, 2025, 61(14): 185-194.

LI Dan, XIA Hongbin, LIU Yuan. Prompt-Enhanced Learning Network for Aspect Sentiment Triplet Extraction[J]. Computer Engineering and Applications, 2025, 61(14): 185-194.

参考文献

[1] SABERI B, SAAD S. Sentiment analysis or opinion mining: a review[J]. International Journal on Advanced Science, Engineering and Information Technology, 2017, 7(5): 1660-1666.
[2] CHEN Z, QIAN T Y. Enhancing aspect term extraction with soft prototypes[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2020: 2107-2117.
[3] WU Z, ZHAO F, DAI X Y, et al. Latent opinions transfer network for target-oriented opinion words extraction[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 9298-9305.
[4] SHARMA T, KAUR K. Aspect sentiment classification using syntactic neighbour based attention network[J]. Journal of King Saud University-Computer and Information Sciences, 2023, 35(2): 612-625.
[5] GAO L, WANG Y L, LIU T C, et al. Question-driven span labeling model for aspect-opinion pair extraction[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2021: 12875-12883.
[6] PENG H Y, XU L, BING L D, et al. Knowing what, how and why: a near complete solution for aspect-based sentiment analysis[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 8600-8607.
[7] XU L, LI H, LU W, et al. Position-aware tagging for aspect sentiment triplet extraction[J]. arXiv:2010.02609, 2020.
[8] WU Z, YING C, ZHAO F, et al. Grid tagging scheme for aspect-oriented fine-grained opinion extraction[J]. arXiv:2010.04640, 2020.
[9] XU L, CHIA Y K, BING L. Learning span-level interactions for aspect sentiment triplet extraction[J]. arXiv:2107.12214, 2021.
[10] YIN Y, WANG C, ZHANG M. PoD: positional dependency-based word embedding for aspect term extraction[J]. arXiv:1911.03785, 2019.
[11] CHEN S W, WANG Y, LIU J, et al. Bidirectional machine reading comprehension for aspect sentiment triplet extraction[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2021: 12666-12674.
[12] YAN H, DAI J Q, JI T, et al. A unified generative framework for aspect?based sentiment analysis[J]. arXiv:2106. 04300, 2021.
[13] ZHANG Y C, YANG Y F, LI Y H, et al. Boundary-driven table-filling for aspect sentiment triplet extraction[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2022: 6485-6498.
[14] BROWN T, MANN B, RYDER N, et al. Language models are few-shot learners[C]//Advances in Neural Information Processing Systems, 2020: 1877-1901.
[15] WANG X L, ZHOU K, WEN J R, et al. Towards unified conversational recommender systems via knowledge-enhanced prompt learning[C]//Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York: ACM, 2022: 1929-1937.
[16] HAN C, WANG Q, CUI Y, et al. E^2VPT: an effective and efficient approach for visual prompt tuning[J]. arXiv:2307. 13770, 2023.
[17] WU H, SHI X D. Adversarial soft prompt tuning for cross-domain sentiment analysis[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2022: 2438-2447.
[18] LI C, GAO F, BU J, et al. SentiPrompt: sentiment knowledge enhanced prompt-tuning for aspect-based sentiment analysis[J]. arXiv:2109.08306, 2021.
[19] CHEN Z, HUANG H, LI B, et al. Semantic and syntactic enhanced aspect sentiment triplet extraction[J]. arXiv:2106. 03315, 2021.
[20] KIPF T, WELLING M. Semi-supervised classification with graph convolutional networks[J]. arXiv:1609.02907, 2016.
[21] CHEN H, ZHAI Z P, FENG F X, et al. Enhanced multi-channel graph convolutional network for aspect sentiment triplet extraction[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics.Stroudsburg: ACL, 2022: 2974-2985.
[22] YANG S H, ZHANG T X, XU H F, et al. Improving aspect sentiment triplet extraction with perturbed masking and edge-enhanced sentiment graph attention network[C]//Proceedings of the International Joint Conference on Neural Networks. Piscataway: IEEE, 2023: 1-8.
[23] PING Z C, SANG G M, LIU Z, et al. Aspect category sentiment analysis based on prompt-based learning with attention mechanism[J]. Neurocomputing, 2024, 565: 126994.
[24] MOVAHEDI S, GHADERY E, FAILI H, et al. Aspect category detection via topic-attention network[J]. arXiv:1901. 01183, 2019.
[25] HE Y, HUANG X Y, ZOU S H, et al. PSAN: prompt semantic augmented network for aspect-based sentiment analysis[J]. Expert Systems with Applications, 2024, 238: 121632.
[26] PENG K, JIANG L, PENG H, et al. Prompt based tri-channel graph convolution neural network for aspect sentiment triplet extraction[C]//Proceedings of the SIAM International Conference on Data Mining, 2024: 145-153.
[27] SHI X F, HU M, REN F J, et al. Prompted representation joint contrastive learning for aspect-based sentiment analysis[J]. Knowledge-Based Systems, 2024, 285: 111345.
[28] DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[J]. arXiv:1810.04805, 2018.
[29] XU Q F, WEI Y W, YUAN S Z, et al. Learning emotional prompt features with multiple views for visual emotion analysis[J]. Information Fusion, 2024, 108: 102366.
[30] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[31] 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.
[32] LI Y, HE Q, ZHANG D. Dual graph convolutional networks integrating affective knowledge and position information for aspect sentiment triplet extraction[J]. Front Neurorobot, 2023, 17: 1193011.
[33] YUAN L, WANG J, YU L C, et al. Encoding syntactic information into transformers for aspect-based sentiment triplet extraction[J]. IEEE Transactions on Affective Computing, 2024, 15(2): 722-735.
[34] WANG L, ZHENG Y, TANG M, et al. Multi-task dual-graph network framework for aspect sentiment triplet extraction[J]. IEEE Access，2022，10：103187-103199.