融合话题信息和语境不协调信息的讽刺识别模型

doi:10.3778/j.issn.1002-8331.2407-0299

摘要/Abstract

摘要： 现有面向话题的讽刺识别研究未同时考虑句内和上下文句间语境不协调信息对讽刺识别的影响，导致讽刺识别的准确度不高。针对这一问题，提出一种融合话题信息和语境不协调信息的讽刺识别模型。使用BERT分别获取话题文本和评论文本的词向量，通过Bi-LSTM进一步提取语义特征，最后使用交叉注意力获取融合话题信息的评论文本特征。此外，将话题文本和评论文本中的名词及名词相邻观点词构成词块，使用Word2Vec获取词块向量，再通过自注意力机制捕捉相同名词不同观点的词块之间的语境不协调信息。将融合话题信息的评论文本特征与词块间语境不协调信息进行拼接，使用Softmax获取讽刺识别结果。实验结果表明，该模型充分考虑了话题信息和语境不协调信息，提高了讽刺识别的准确率。

关键词: 讽刺识别, 面向话题的讽刺识别, 不协调信息, 注意力机制

Abstract: Existing topic-oriented sarcasm detection studies have not simultaneously considered the effects of intra-sentence and inter-contextual sentence contextual incongruity information on sarcasm detection, resulting in poor sarcasm recognition accuracy. To address this problem, a sarcasm detection model incorporating topic information and contextual incongruity information is proposed. Firstly, the word vectors of topic text and comment text are obtained separately using BERT, then the semantic features are further extracted using Bi-LSTM, and finally the comment text features fused with topic information are obtained using cross-attention. In addition, the nouns and noun-adjacent opinion words in the topic text and comment text are formed into word blocks, and Word2Vec is used to obtain the word block vectors, and then the self-attention mechanism is used to capture the contextual incongruity information between the word blocks of the same nouns with different opinions. The comment text features fused with topic information are spliced with the context incongruity information between word blocks, and then Softmax is used to obtain the sarcasm detection results. The experimental results show that the model fully considers the topic information and context incongruity information to improve the accuracy of sarcasm detection.

Key words: sarcasm detection, topic-oriented sarcasm detection, incongruity information, attention mechanism

段玉俊, 张顺香, 钱龙海, 文华, 丁远远, 葛唱. 融合话题信息和语境不协调信息的讽刺识别模型[J]. 计算机工程与应用, 2025, 61(17): 251-258.

DUAN Yujun, ZHANG Shunxiang, QIAN Longhai, WEN Hua, DING Yuanyuan, GE Chang. Sarcasm Detection Model Incorporating Topic Information and Contextual Incongruity Information[J]. Computer Engineering and Applications, 2025, 61(17): 251-258.

参考文献

[1] 梁斌, 林子杰, 徐睿峰, 等. 面向话题的讽刺识别: 新任务、新数据和新方法[J]. 中文信息学报, 2023, 37(2): 138-147.
LIANG B, LIN Z J, XU R F, et al. Topic-oriented sarcasm detection: new task, new dataset and new methond[J]. Journal of Chinese Information Processing, 2023, 37(2): 138-147.
[2] REN Y F, WANG Z L, PENG Q, et al. A knowledge-augmented neural network model for sarcasm detection[J]. Information Processing & Management, 2023, 60(6): 103521.
[3] HE Y L, CHEN M J, HE Y Y, et al. Sarcasm detection base on adaptive incongruity extraction network and incongruity cross-attention[J]. Applied Sciences, 2023, 13(4): 2102.
[4] PAN H, LIN Z, FU P, et al. Modeling the incongruity between sentence snippets for sarcasm detection[C]//Findings of the Association for Computational Linguistics: EMNLP, 2020: 2132-2139.
[5] WANG X B, DONG Y Q, JIN D, et al. Augmenting affective dependency graph via iterative incongruity graph learning for sarcasm detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2023: 4702-4710.
[6] SREELAKSHMI K, RAFEEQUE P C. An effective approach for detection of sarcasm in tweets[C]//Proceedings of the International CET Conference on Control, Communication, and Computing. Piscataway: IEEE, 2018: 377-382.
[7] HUANG P L, ZENG X W, WENG J T, et al. SICKNet: a humor detection network integrating semantic incongruity and commonsense knowledge[C]//Proceedings of the IEEE 34th International Conference on Tools with Artificial Intelligence. Piscataway: IEEE, 2022: 288-296.
[8] RILOFF E, QADIR A, SURVE P, et al. Sarcasm as contrast between a positive sentiment and negative situation[C]//Proceedings of the Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2013: 704-714.
[9] MAYNARD D, GREENWOOD M. Who cares about sarcastic tweets? investigating the impact of sarcasm on sentiment analysis[C]//Proceedings of the International Conference on Language Resources and Evaluation, 2014.
[10] BHARTI S K, BABU K S, JENA S K. Parsing-based sarcasm sentiment recognition in twitter data[C]//Proceedings of the IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining. New York: ACM, 2015: 1373-1380.
[11] GOVINDAN V, BALAKRISHNAN V. A machine learning approach in analysing the effect of hyperboles using negative sentiment tweets for sarcasm detection[J]. Journal of King Saud University-Computer and Information Sciences, 2022, 34(8): 5110-5120.
[12] 韦斯羽, 朱广丽, 谈光璞, 等. 融合反讽语言特征的反讽语句识别模型[J]. 智能系统学报, 2024, 19(3): 689-696.
WEI S Y, ZHU G L, TAN G P, et al. Ironic sentence recognition model integrating ironic language features[J]. CAAI Transactions on Intelligent Systems, 2024, 19(3): 689-696.
[13] VINOTH D, PRABHAVATHY P. An intelligent machine learning-based sarcasm detection and classification model on social networks[J]. The Journal of Supercomputing, 2022, 78(8): 10575-10594.
[14] PANDEY R, SINGH J P. BERT-LSTM model for sarcasm detection in code-mixed social media post[J]. Journal of Intelligent Information Systems, 2023, 60(1): 235-254.
[15] ZHANG Y Z, MA D, TIWARI P, et al. Stance-level sarcasm detection with BERT and stance-centered graph attention networks[J]. ACM Transactions on Internet Technology, 2023, 23(2): 1-21.
[16] 樊小超, 杨亮, 林鸿飞, 等. 基于多语义融合的反讽识别[J]. 中文信息学报, 2021, 35(6): 103-111.
FAN X C, YANG L, LIN H F, et al. Irony recognition based on multiple semantic fusion[J]. Journal of Chinese Information Processing, 2021, 35(6): 103-111.
[17] WEN Z Y, GUI L, WANG Q L, et al. Sememe knowledge and auxiliary information enhanced approach for sarcasm detection[J]. Information Processing & Management, 2022, 59(3): 102883.
[18] ZHANG S X, ZHU A Q, ZHU G L, et al. Building fake review detection model based on sentiment intensity and PU lear-ning[J]. IEEE Transactions on Neural Networks and Lear-ning Systems, 2023, 34(10): 6926-6939.
[19] DU Y, LI T, PATHAN M S, et al. An effective sarcasm detection approach based on sentimental context and individual expression habits[J]. Cognitive Computation, 2022, 14(1): 78-90.
[20] 魏鹏飞, 曾碧, 廖文雄. 基于RoBERTa的社交媒体会话中的讽刺检测模型[J]. 计算机工程与应用, 2022, 58(13): 164-170.
WEI P F, ZENG B, LIAO W X. RoBERTa-based sarcasm detection model in conversation threads from social media[J]. Computer Engineering and Applications, 2022, 58(13): 164-170.
[21] LI J N, PAN H L, LIN Z, et al. Sarcasm detection with commonsense knowledge[J]. ACM Transactions on Audio, Speech, and Language Processing, 2021, 29: 3192-3201.
[22] MENG J N, ZHU Y L, SUN S C, et al. Sarcasm detection based on BERT and attention mechanism[J]. Multimedia Tools and Applications, 2024, 83(10): 29159-29178.
[23] CHEN W Q, LIN F Q, ZHANG X, et al. Jointly learning sentimental clues and context incongruity for sarcasm detection[J]. IEEE Access, 2022, 10: 48292-48300.
[24] DEVLIN J, CHANG M W, LEE K, et al. BERT: PRE-training of deep bidirectional transformers for language understa-nding[J]. arXiv:1810.04805, 2018.
[25] ZHOU P, SHI W, TIAN J, et al. Attention-based bidirectional long short-term memory networks for relation classification[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2016: 207-212.
[26] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.
[27] MIKOLOV T, CHEN K, CORRADO G, et al. Efficient estimation of word representations in vector space[J]. arXiv:1301. 3781, 2013.
[28] LOU C W, LIANG B, GUI L, et al. Affective dependency graph for sarcasm detection[C]//Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York: ACM, 2021: 1844-1849.