语义信息提取和图结构挖掘的事件骨架生成方法

doi:10.3778/j.issn.1002-8331.2403-0364

摘要/Abstract

摘要： 事件骨架生成旨在从一系列的事件图中归纳出包含事件类型及其时序关系的事件骨架图。这是在时间复杂事件模式归纳任务中的一个核心步骤。尽管现有的方法在这项任务上已经取得了一定的效果，但是由于事件图的复杂性和多变性，这些方法在挖掘事件图的结构信息和语义信息方面仍显不足。因此，为解决该问题，提出了一种事件骨架生成模型。在图编码阶段，模型使用了拉普拉斯位置编码，以精准捕捉和编码图结构的局部信息。同时，模型采用了多头注意力机制和图卷积网络，以提取语义信息和图结构信息，全面总结事件发展的全局结构信息，构建出更泛化、更全面的事件骨架图。实验证明，在事件骨架生成任务上，该模型在Event Match指标上提升了8.83%，Event Sequence Match指标上提升了11.2%（L=2）和7.6%（L=3），实现了较大的性能提升。

关键词: 事件模式归纳, 事件骨架生成, 图生成, 语义信息提取, 图结构挖掘

Abstract: The event skeleton generation aims to induce an event skeleton graph containing event types and their temporal relationships from a series of event graphs. This is a core step in the task of temporal complex event schema induction. Although existing methods have achieved certain effectiveness in this task, they still lack in mining structural and semantic information from event graphs due to their complexity and variability. Therefore, to address this issue, a novel event skeleton generation model is proposed in this paper. In the graph encoding stage, the model utilizes Laplacian positional encoding to accurately capture and encode the local information of the graph structure. Meanwhile, the model adopts multi-head attention mechanisms and graph convolutional networks to extract semantic and structural information, comprehensively summarizing the global structural information of event development and constructing more generalized and comprehensive event skeleton graphs. Experimental results demonstrate that in the event skeleton generation task, the model achieves an improvement of 8.83% on the Event Match metric and improvements of 11.2% (L=2) and 7.6% (L=3) on the Event Sequence Match metric, achieving significant performance gains.

Key words: event schema induction, event skeleton generation, graph generation, semantic information extraction, graph structure mining

黄凯, 马廷淮, 孙圣杰, 龚智恒, 汤毅翔, 陈思. 语义信息提取和图结构挖掘的事件骨架生成方法[J]. 计算机工程与应用, 2025, 61(12): 187-195.

HUANG Kai, MA Tinghuai, SUN Shengjie, GONG Zhiheng, TANG Yixiang, CHEN Si. Event Skeleton Generation Method for Semantic Information Extraction and Graph Structure Mining[J]. Computer Engineering and Applications, 2025, 61(12): 187-195.

参考文献

[1] 刘玉婷, 丁鲲, 刘茗, 等. 结合对抗训练增强和联合损失微调的脚本事件预测方法[J]. 小型微型计算机系统, 2025, 46(2): 274-279.
LIU Y T, DING K, LIU M, et al. Script event prediction based on adversarial training augmentation and joint loss fine-tuning[J]. Journal of Chinese Computer Systems, 2025, 46(2): 274-279.
[2] 单晓红, 庞世红, 刘晓燕, 等. 基于事理图谱的网络舆情事件预测方法研究[J]. 情报理论与实践, 2020, 43(10): 165-170.
SHAN X H, PANG S H, LIU X Y, et al. Research on Internet public opinion event prediction method based on event evolution graph[J]. Information Studies (Theory & Application), 2020, 43(10): 165-170.
[3] MA T H, RONG H, HAO Y S, et al. A novel sentiment polarity detection framework for Chinese[J]. IEEE Transactions on Affective Computing, 2022, 13(1): 60-74.
[4] ZHOU H H, MA T H, RONG H, et al. MDMN: multi-task and domain adaptation based multi-modal network for early rumor detection[J]. Expert Systems with Applications, 2022, 195: 116517.
[5] LI M L, LI S, WANG Z, et al. The future is not one-dimensional: complex event schema induction by graph modeling for event prediction[C]//Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2021: 5203-5215.
[6] JIN X M, LI M L, JI H. Event schema induction with double graph autoencoders[C]//Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2022: 2013-2025.
[7] CHAMBERS N. Event schema induction with a probabilistic entity-driven model[C]//Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2013: 1797-1807.
[8] CHEUNG J C K, POON H, VANDERWENDE L. Probabilistic frame induction[C]//Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2013: 837-846.
[9] NGUYEN K H, TANNIER X, FERRET O, et al. Generative event schema induction with entity disambiguation[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2015: 188-197.
[10] SHA L, LI S J, CHANG B B, et al. Joint learning templates and slots for event schema induction[C]//Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2016: 428-434.
[11] HUANG L F, CASSIDY T, FENG X C, et al. Liberal event extraction and event schema induction[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2016: 258-268.
[12] YUAN Q, REN X, HE W Q, et al. Open-schema event profiling for massive news corpora[C]//Proceedings of the 27th ACM International Conference on Information and Knowledge Management. New York: ACM, 2018: 587-596.
[13] LIN Y, JI H, HUANG F, et al. A joint neural model for information extraction with global features[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2020: 7999-8009.
[14] CHAMBERS N, JURAFSKY D. Unsupervised learning of narrative event chains[C]//Proceedings of ACL-08: HLT, 2008: 789-797.
[15] CHAMBERS N, JURAFSKY D. Unsupervised learning of narrative schemas and their participants[C]//Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: Volume 2-ACL-IJCNLP’09. Morristown: ACL, 2009: 602.
[16] RUDINGER R, RASTOGI P, FERRARO F, et al. Script induction as language modeling[C]//Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2015: 1681-1686.
[17] PICHOTTA K, MOONEY R. Learning statistical scripts with LSTM recurrent neural networks[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2016.
[18] WEBER N, BALASUBRAMANIAN N, CHAMBERS N. Event representations with tensor-based compositions[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2018.
[19] WEBER N, RUDINGER R, VAN DURME B. Causal inference of script knowledge[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2020: 7583-7596.
[20] MOSTAFAZADEH N, CHAMBERS N, HE X D, et al. A corpus and cloze evaluation for deeper understanding of commonsense stories[C]//Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2016: 839-849.
[21] WEBER N, SHEKHAR L, KWON H, et al. Generating narrative text in a switching dynamical system[C]//Proceedings of the 24th Conference on Computational Natural Language Learning. Stroudsburg: ACL, 2020: 520-530.
[22] KWON H, KOUPAEE M, SINGH P, et al. Modeling preconditions in text with a crowd-sourced dataset[C]//Findings of the Association for Computational Linguistics: EMNLP 2020. Stroudsburg: ACL, 2020: 3818-3828.
[23] LI M L, ZENG Q, LIN Y, et al. Connecting the dots: event graph schema induction with path language modeling[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2020: 684-695.
[24] DU X Y, ZHANG Z X, LI S, et al. RESIN-11: schema-guided event prediction for 11 newsworthy scenarios[C]//Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies: System Demonstrations. Stroudsburg: ACL, 2022: 54-63.
[25] KREUZER D, BEAINI D, HAMILTON W L, et al. Rethinking graph transformers with spectral attention[J]. arXiv:2106. 03893, 2021.
[26] ZHANG M, JIANG S, CUI Z, et al. D-VAE: a variational autoencoder for directed acyclic graphs[C]//Advances in Neural Information Processing Systems, 2019.