结合对比学习和迭代优化的事件类型归纳方法

doi:10.3778/j.issn.1002-8331.2309-0339

摘要/Abstract

摘要： 事件类型归纳能够从无标注文本中自动发现并命名新事件类型，可以有效获取多个领域的事件知识。现有研究将所有样本视为单一事件样本，仅考虑样本包含的某个事件类型，忽略了多事件样本对事件语义学习和事件类型命名的负面影响。针对上述问题，提出了一种结合对比学习和迭代优化的事件类型归纳方法。针对多事件样本对事件语义学习的影响，提出了一种基于提示学习的多事件检测方法，在模型训练前检测并剔除多事件样本。为了优化事件语义表示，提出了一种基于抽象语义表示（abstract meaning representation，AMR）的候选触发词识别策略，并引入外部锚点和聚类伪标签，优化对比学习训练效果。为了提升未知事件类型的命名质量，提出了一种基于ChatGPT反馈的事件类型命名迭代优化方法，根据ChatGPT的命名结果，剔除影响事件类型命名的样本，并使用经过处理的数据集微调模型。迭代上述过程，直到生成预期质量的事件类型名称。在ACE2005数据集上的实验结果表明，该方法能够显著提升未知事件类型的聚类效果，并能够有效生成高质量的事件类型名称。

关键词: 事件类型归纳, 对比学习, 大型语言模型

Abstract: Event type induction is the task of automatically discovering and naming new event types from unlabeled text, enabling the acquisition of event knowledge across multiple domains. Existing research treats all samples as single-event samples, considering only the specific event type contained within each sample, thus overlooking the negative impact of multi-event samples on event semantic learning and event type naming. To address these issues, this paper proposes a new method combined with contrastive learning and iterative optimization. Firstly, to mitigate the influence of multi-event samples on event semantic learning, a prompt-based multi-event detection method is introduced to detect and exclude multi-event samples prior to model training. Secondly, to optimize event semantic representations, a candidate trigger word recognition strategy based on abstract meaning representation (AMR) is proposed, along with the incorporation of external anchors and cluster pseudo-labels to enhance the effectiveness of contrastive learning training. Finally, to improve the quality of naming unknown event types, an iterative optimization method for event type naming is proposed. Based on the naming results from ChatGPT, samples that affect event type naming are filtered out, and the model is fine-tuned using the processed dataset. This iterative process continues until event type names of the desired quality are generated. Experimental results on the ACE2005 dataset show that the proposed method significantly enhances the clustering performance of unknown event types and effectively generates high-quality event type names.

Key words: event type induction, contrastive learning, large language model

冀义豪, 任一支, 袁理锋, 刘容轲, 潘高宁. 结合对比学习和迭代优化的事件类型归纳方法[J]. 计算机工程与应用, 2025, 61(3): 196-211.

JI Yihao, REN Yizhi, YUAN Lifeng, LIU Rongke, PAN Gaoning. Event Type Induction Combined with Contrastive Learning and Iterative Optimization[J]. Computer Engineering and Applications, 2025, 61(3): 196-211.

参考文献

[1] CHEN Y, XU L, LIU K, et al. Event extraction via dynamic multi-pooling convolutional neural networks[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, 2015: 167-176.
[2] LIU S, CHEN Y, LIU K, et al. Exploiting argument information to improve event detection via supervised attention mechanisms[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, 2017: 1789-1798.
[3] YANG S, FENG D, QIAO L, et al. Exploring pre-trained language models for event extraction and generation[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019: 5284-5294.
[4] DU X, CARDIE C. Event extraction by answering (almost) natural questions[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing(EMNLP), 2020: 671-683.
[5] MA Y, WANG Z, CAO Y, et al. Prompt for extraction? PAIE: prompting argument interaction for event argument extraction[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2022: 6759-6774.
[6] HUANG L, JI H, CHO K, et al. Zero-shot transfer learning for event extraction[C]//Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2018: 2160-2170.
[7] ZHANG H, WANG H, ROTH D. Zero-shot label-aware event trigger and argument classification[C]//Findings of the Association for Computational Linguistics 2021, 2021: 1331-1340.
[8] LYU Q, ZHANG H, SULEM E, et al. Zero-shot event extraction via transfer learning: challenges and insights[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, 2021: 322-332.
[9] ZHANG S, JI T, JI W, et al. Zero-shot event detection based on ordered contrastive learning and prompt-based prediction[C]//Findings of the Association for Computational Linguistics: NAACL 2022, 2022: 2572-2580.
[10] LI S, JI H, HAN J. Open relation and event type discovery with type abstraction[C]//Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, 2022: 6864-6877.
[11] HUANG L, CASSIDY T, FENG X, et al. Liberal event extraction and event schema induction[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (ACL 2016), 2016: 258-268.
[12] HUANG L, JI H. Semi-supervised new event type induction and event detection[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020: 718-724.
[13] SHEN J, ZHANG Y, JI H, et al. Corpus-based open-domain event type induction[C]//Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 2021: 5427-5440.
[14] CAO P, HAO Y, CHEN Y, et al. Event ontology completion with hierarchical structure evolution networks[C]//Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023: 306-320.
[15] ZHAO J, GUI T, ZHANG Q, et al. A relation-oriented clustering method for open relation extraction[C]//Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 2021: 9707-9718.
[16] WALKER C, STRASSEL S, MEDERO J, et al. ACE 2005 multilingual training corpus ldc2006t06, 2006[EB/OL]. [2006]. https://catalog.ldc.upenn.edu/LDC2006T06.
[17] GRISHMAN R. Information extraction: capabilities and challenges[R]. 2012 International Winter School in Language and Speech Technologies, 2012.
[18] 刘泽旖, 余文华, 洪智勇, 等. 基于问题回答模式的中文事件抽取[J]. 计算机工程与应用, 2023, 59(2): 153-160.
LIU Z Y, YU W H, HONG Z Y, et al. Chinese event extraction using question answering[J]. Computer Engineering and Applications, 2023, 59(2): 153-160.
[19] RILOFF E. Automatically constructing a dictionary for information extraction tasks[C]//Proceedings of the Eleventh National Conference on Artificial Intelligence, 1993: 811-816.
[20] YAKUSHIJI A, TATEISI Y, MIYAO Y, et al. Event extraction from biomedical papers using a full parser[C]//Proceedings of the Pacific Symposium on Biocomputing, Pacific Symposium on Biocomputing, 2001: 408-419.
[21] DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[C]//Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), 2019: 4171-4186.
[22] RADFORD A, WU J, CHILD R, et al. Language models are unsupervised multitask learners[EB/OL]. [2023-09-01]. https://gwerb.net/doc/ai/nn/transformer/gpt/2019-radford. pdf.
[23] 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.
[24] GAO T, FISCH A, CHEN D. Making pre-trained language models better few-shot learners[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), 2021: 3816-3830.
[25] SCHICK T, SCHüTZE H. It’s not just size that matters: small language models are also few-shot learners[C]//Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021: 2339-2352.
[26] 鲍彤, 章成志. ChatGPT中文信息抽取能力测评——以三种典型的抽取任务为例[J]. 数据分析与知识发现, ?2023, 7(9): 1-11.
BAO T, ZHANG C Z. Extracting Chinese information with ChatGPT: an empirical study by three typical tasks[J]. Data Analysis and Knowledge Discovery, 2023, 7(9): 1-11.
[27] 王泽深, 杨云, 向鸿鑫, 等. 零样本学习综述[J]. 计算机工程与应用, 2021, 57(19): 1-17.
WANG Z S, YANG Y, XIANG H X, et al. Survey on zero-shot learning[J]. Computer Engineering and Applications, 2021, 57(19): 1-17.
[28] 张森辉. 基于有序对比学习的零样本事件检测技术[D]. 上海: 华东师范大学, 2023.
ZHANG S H. Zero-shot event detection based on ordered contrastive learning[D]. Shanghai: East China Normal University, 2023.
[29] XIAN Y, AKATA Z, SHARMA G, et al. Latent embeddings for zero-shot classification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 69-77.
[30] CHEN Z, LI J, LUO Y, et al. CANZSL: cycle-consistent adversarial networks for zero-shot learning from natural language[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2020: 874-883.
[31] 赵鹏, 汪纯燕, 张思颖, 等. 一种基于融合重构的子空间学习的零样本图像分类方法[J]. 计算机学报, 2021, 44(2): 409-421.
ZHAO P, WANG C Y, ZHANG S Y, et al. A Zero-shot image classification method based on subspace learning with the fusion of reconstruction[J]. Chinese Joural of Computers, 2021, 44(2): 409-421.
[32] XIAN Y, LORENZ T, SCHIELE B, et al. Feature generating networks for zero-shot learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 5542-5551.
[33] FENG Y, HUANG X, YANG P, et al. Non-generative generalized zero-shot learning via task-correlated disentanglement and controllable samples synthesis[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022: 9346-9355.
[34] 张海涛, 苏琳. 结合知识图谱的变分自编码器零样本图像识别[J]. 计算机工程与应用, 2023, 59(1): 236-243.
ZHANG H T, SU L. Variational auto-encoder combined with knowledge graph zero-shot learning[J]. Computer Engineering and Applications, 2023, 59(1): 236-243.
[35] CHEN T, KORNBLITH S, NOROUZi M, et al. A simple framework for contrastive learning of visual representations[C]//Proceedings of the International Conference on Machine Learning, 2020: 1597-1607.
[36] FANG K, XIE P. 2020. Cert: contrastive selfsupervised learning for language understanding[J]. arXiv:2005.12766, 2020.
[37] GAO T, YAO X, CHEN D. SimCSE: simple contrastive learning of sentence embeddings[C]//Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP 2021), 2021: 6894-6910.
[38] ZHANG D, NAN F, WEI X, et al. Supporting clustering with contrastive learning[C]//Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021: 5419-5430.
[39] LIANG B, ZHU Q, LI X, et al. Jointcl: a joint contrastive learning framework for zero-shot stance detection[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2022: 81-91.
[40] LI J, SHANG J, MCAULEY J. UCTopic: unsupervised contrastive learning for phrase representations and topic mining[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volme 1: Long Papers), 2022: 6159-6169.
[41] OpenAI. ChatGPT: optimizing language models for dialogue[EB/OL]. [2023-01-12]. https://openai. com/blog/chatgpt/.
[42] ZENG A, LIU X, DU Z, et al. GLM-130B: an open bilingual pre-trained model[C]//Proceedings of the Eleventh International Conference on Learning Representations, 2022.
[43] VAN DER MAATEN L, HINTON G. Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008, 9(11): 2579-2605.