中文突发事件论元识别方法研究
Research on the Method of Chinese Emergency Event Argument Recognition

DOI: 10.12677/CSA.2022.122041, PP. 405-411

Keywords: 事件论元识别，双向长短期记忆神经网络，图注意力机制
Event Argument Recognition, BiSTM, Graph Attention Network

Full-Text   Cite this paper   Add to My Lib

Abstract:

事件抽取是从非结构化的文本中抽取用户感兴趣的事件信息，并以结构化的形式展现。当前社交媒体快速发展，互联网上突发事件信息数量也剧增。如何准确地从大量无结构事件信息中识别并抽取出突发事件信息，分析突发事件舆情趋势对社会安全极为重要。本文利用双向长短期记忆神经网络(BiLSTM)与图注意力机制(GAT)获取事件句子中的句法信息和获得事件论元之间的内在关联，进一步提升事件论元识别的准确率。通过在真实数据集中多次实验证明，本文中的方法在公开的数据集上进行验证，与以往的事件论元识别方法相比获得较大的性能提升。
Event extraction is to extract the event information that users are interested in from unstructured text and display it in a structured form. With the rapid development of social media, the amount of emergency event information on the Internet has also increased dramatically. How to accurately identify and extract emergent event information from a large amount of unstructured event information, and analyze the trend of public opinion on emergencies is extremely important to social security. In this paper, the Bi-directional Long Short-Term Memory (BiLSTM) and the graph attention network (GAT) are used to obtain the syntactic information in the event sentence and obtain the intrinsic correlation between the event arguments, so as to further improve the accuracy of the event argument recognition. Multiple experiments on real datasets prove that the method in this paper is validated on public datasets and achieves a large performance improvement compared with previous event argument extraction methods.

References

[1]	Doddington, G., Mitchell, A., Przybocki, M., et al. (2004) The Automatic Content Extraction (ACE) Program-Tasks, Data, and Evaluation. Proceedings of the 2004 International Conference on Language Resources and Evaluation, Lisbon, September 2004, 837-840.
[2]	高李政, 周刚, 罗军勇, 兰明敬. 元事件抽取研究综述[J]. 计算机科学, 2019, 46(8): 9-15.
[3]	Ahn, D. (2006) The Stages of Event Extraction. Proceddings of the Workshop on Annotating and Reasoning about Time and Events, Sydney, July 2006, 1-8.
[4]	McClosky, D., Surdeanu, M. and Manning, C.D. (2011) Event Extraction as Dependency Parsing. Proceedings of the 49th Annual Meeting of the Association for Compu-tational Linguistics: Human Language Technologies, Portland, 1 June 2011, 1626-1635.
[5]	Li, P.F., Zhu, Q.M. and Zhou. G.D. (2013) Joint Modeling of Argument Identification and Role Determination in Chinese Event Extraction with Discourse-Level Information. Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence, Suzhou, 3 August 2013, 2120-2126.
[6]	秦彦霞, 张民, 郑德权. 神经网络事件抽取技术综述[J]. 智能计算机与应用, 2018, 8(3): 1-5+10.
[7]	Chen, Y., Xu, L., Liu, K., et al. (2015) Event Extraction via Dynamic Multi-Pooling Convolutional Neural Networks. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguis-tics and the 7th International Joint Conference on Natural Language Processing, Beijing, July 2015, 167-176. https://doi.org/10.3115/v1/P15-1017
[8]	Sha, L., Qian, F., Chang, B.B. and Sui, Z.F. (2018) Jointly Extraction Event Triggers and Arguments by Dependency-Bridge RNN and Tensor-Based Argument Interaction. https://aaai.org/ocs/index.php/AAAI/AAAI18/paper/view/16222/16157
[9]	Lin, H.Y., Lu, Y.J., Han, X.P., Sun, L. (2016) A Convolution Bilstm Neural Network Model for Chinese Event Extraction. https://eprints.lancs.ac.uk/id/eprint/83783/1/160.pdf
[10]	Yang, S., Feng, D.W., Qiao, L.B., Kan, Z.G. and Li, D.S. (2019) Exploring Pre-Trained Language Models for Event Extraction and Geenration. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, July 2019, 5284-5294. https://doi.org/10.18653/v1/P19-1522
[11]	Zheng, S., Cao, W., Xu, W. and Bian, J. (2019) Doc2EDAG: An End-to-End Document-Level Framework for Chinese Financial Event Extraction. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), Hong Kong, November 2019, 337-346. https://doi.org/10.18653/v1/D19-1032

Full-Text