基于多提示学习的方面类别情感分析方法

doi:10.3778/j.issn.1673-9418.2407093

摘要/Abstract

摘要： 基于方面类别的情感分析（ACSA）旨在辨别评论文本中的方面类别，并同时预测它们的情感极性，是情感分析领域重要的细粒度子任务。近年来，基于预训练语言模型的微调（Fine-tuning）方法已经为方面类别情感分析提供了有效的解决思路。然而，由于预训练任务和下游情感分类任务目标不一致，影响了情感分析质量提升的空间。目前基于提示模板的提示学习（Prompt learning）能够对其进行相应缓解，但人工设计单一的Prompt文本为ACSA任务提供的上下文有限，缺少丰富性。针对此问题，提出了一种基于提示学习的方面类别情感分析方法（Multi-Prompt_ACSA）。在提示学习的基础上进行了提示模板工程和答案工程的多样化设计，结合ACSA的研究特点，提出了适配方面类别情感分析的提示学习方法。引入自回归预训练语言模型进行训练。基于Prompt的多样化设计集成多个不同提示模板下的情感分类结果。与其他模型（非预训练、预训练和提示学习三个类别）在SemEval 2015和SemEval 2016数据集上的结果相比，提出的方法在F1指标上有良好的效果提升。

关键词: 方面类别情感分析, 提示学习, Prompt多样化设计

Abstract: Aspect-category sentiment analysis (ACSA) aims to discern aspect categories in review texts and simultaneously predict their sentiment polarity. It is an important fine-grained subtask in the field of sentiment analysis. Currently, fine-tuning with pretrained language models shows effectiveness in ACSA, but its training tasks are different from downstream ACSA, which limits its analysis quality. Although the prompt template based prompt learning shows its good performance, it is not diverse enough to cover different contexts in the manually designed prompts for ACSA. To solve this problem, this paper proposes an aspect category sentiment analysis method (Multi-Prompt_ACSA) based on prompt learning. Firstly, on the basis of prompt learning, the diversified design of prompt template engineering and answer engineering is carried out. Based on the characteristics of ACSA task, a prompt learning method is proposed to match aspect-category sentiment analysis. Then, an autoregressive pretrained language model is introduced. Further, classification results are integrated based on Prompt??s diverse design. Compared with the results of other models including non-pretrain, pretrain and prompt learning, on the SemEval 2015 and SemEval 2016 datasets, the proposed prompt-based learning method has fine improvement in terms of F1.

Key words: aspect-category sentiment analysis, prompt learning, diversified Prompt design

刘锦行, 李琳, 吴任伟, 刘佳. 基于多提示学习的方面类别情感分析方法[J]. 计算机科学与探索, 2025, 19(5): 1334-1341.

LIU Jinhang, LI Lin, WU Renwei, LIU Jia. Multi-prompt Learning Based Aspect-Category Sentiment Analysis[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(5): 1334-1341.

参考文献

[1] CAI H J, TU Y F, ZHOU X S, et al. Aspect-category based sentiment analysis with hierarchical graph convolutional network[C]//Proceedings of the 28th International Conference on Computational Linguistics. Stroudsburg: ACL, 2020: 833-843.
[2] SCHMITT M, STEINHEBER S, SCHREIBER K, et al. Joint aspect and polarity classification for aspect-based sentiment analysis with end-to-end neural networks[C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2018: 1109-1114.
[3] LIANG B, SU H, YIN R D, et al. Beta distribution guided aspect-aware graph for aspect category sentiment analysis with affective knowledge[C]//Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2021: 208-218.
[4] DAI Z H, DAI W, LIU Z H, et al. Multi-task multi-head attention memory network for fine-grained sentiment analysis[C]//Proceedings of the 8th CCF International Conference on Natural Language Processing and Chinese Computing. Cham: Springer, 2019: 609-620.
[5] GUO X, ZHANG G, WANG S G, et al. Multi-way matching based fine-grained sentiment analysis for user reviews[J]. Neural Computing and Applications, 2020, 32(10): 5409-5423.
[6] FU Y J, LIAO J, LI Y, et al. Multiple perspective attention based on double BiLSTM for aspect and sentiment pair extract[J]. Neurocomputing, 2021, 438: 302-311.
[7] HU M T, ZHAO S W, ZHANG L, et al. CAN: constrained attention networks for multi-aspect sentiment analysis[EB/OL]. [2024-04-23]. https://arxiv.org/abs/1812.10735.
[8] 孙小婉, 王英, 王鑫, 等. 面向双注意力网络的特定方面情感分析模型[J]. 计算机研究与发展, 2019, 56(11): 2384-2395.
SUN X W, WANG Y, WANG X, et al. Aspect-based sentiment analysis model based on dual-attention networks[J]. Journal of Computer Research and Development, 2019, 56(11): 2384-2395.
[9] 张文轩, 殷雁君, 智敏. 用于方面级情感分析的情感增强双图卷积网络[J]. 计算机科学与探索, 2024, 18(1): 217-230.
ZHANG W X, YIN Y J, ZHI M. Affection enhanced dual graph convolution network for aspect based sentiment analysis[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(1): 217-230.
[10] XUE W, LI T. Aspect based sentiment analysis with gated convolutional networks[EB/OL]. [2024-04-23]. https://arxiv.org/abs/1805.07043.
[11] LIU P F, YUAN W Z, FU J L, et al. Pre-train, prompt, and predict: a systematic survey of prompting methods in natural language processing[J]. ACM Computing Surveys, 2023, 55(9): 1-35.
[12] JIANG Q, CHEN L T, XU R F, et al. A challenge dataset and effective models for aspect-based sentiment analysis[C]//Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2019: 6280-6285.
[13] YIN R D, SU H, LIANG B, et al. Extracting the collaboration of entity and attribute: gated interactive networks for aspect sentiment analysis[C]//Proceedings of the 9th CCF International Conference on Natural Language Processing and Chinese Computing. Cham: Springer, 2020: 802-814.
[14] LI Y C, YIN C X, ZHONG S H, et al. Multi-instance multi-label learning networks for aspect-category sentiment analysis[EB/OL]. [2024-04-23]. https://arxiv.org/abs/2010.02656.
[15] SCHICK T, SCHüTZE H. Exploiting cloze-questions for few-shot text classification and natural language inference[C]//Proceedings of the 2022 Conference of the European Chapter of the Association for Computational Linguistics. Stroudsburg: ACL, 2022: 255-269.
[16] CHEN X, ZHANG N Y, XIE X, et al. KnowPrompt: knowledge-aware prompt-tuning with synergistic optimization for relation extraction[C]//Proceedings of the ACM Web Conference 2022. New York: ACM, 2022: 2778-2788.
[17] SCHICK T, SCHMID H, SCHüTZE H. Automatically identifying words that can serve as labels for few-shot text classification[EB/OL]. [2024-05-14]. https://arxiv.org/abs/2010. 13641.
[18] LIU X, ZHENG Y N, DU Z X, et al. GPT understands, too[J]. AI Open, 2024, 5: 208-215.
[19] GAO T Y, FISCH A, CHEN D Q. Making pre-trained language models better few-shot learners[EB/OL]. [2024-05-14]. https://arxiv.org/abs/2012.15723.
[20] HU S D, DING N, WANG H D, et al. Knowledgeable prompt-tuning: incorporating knowledge into prompt verbalizer for text classification[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2022: 2225-2240.
[21] BROWN T B, MANN B, RYDER N, et al. Language models are few-shot learners[C]//Proceedings of the 34th International Conference on Neural Information Processing Systems, 2020: 1877-1901.
[22] SHIN T, RAZEGHI Y, LOGAN R L, et al. AutoPrompt: eliciting knowledge from language models with automatically generated prompts[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2020: 4222-4235.
[23] YANG Z L, DAI Z H, YANG Y M, et al. XLNet: generalized autoregressive pretraining for language understanding[C]//Advances in Neural Information Processing Systems 32, 2019: 5754-5764.
[24] DAI Z H, YANG Z L, YANG Y M, et al. Transformer-XL: attentive language models beyond a fixed-length context[EB/OL]. [2024-05-14]. https://arxiv.org/abs/1901.02860.
[25] LI Y F, GUO L Z, ZHOU Z H. Towards safe weakly supervised learning[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(1): 334-346.
[26] 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.
[27] 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.