计算机科学 ›› 2019, Vol. 46 ›› Issue (9): 201-205.doi: 10.11896/j.issn.1002-137X.2019.09.029

• 人工智能 • 上一篇    下一篇

基于注意力机制的事件同指消解方法

程昊熠, 李培峰, 朱巧明   

  1. (苏州大学计算机科学与技术学院 江苏 苏州215006);
    (江苏省计算机信息技术处理重点实验室 江苏 苏州215006)
  • 收稿日期:2018-08-24 出版日期:2019-09-15 发布日期:2019-09-02
  • 通讯作者: 李培峰(1971-),男,教授,硕士生导师,主要研究领域为自然语言处理和机器学习,E-mail:pfli@suda.edu.cn
  • 作者简介:程昊熠(1995-),男,硕士生,主要研究领域为自然语言处理;朱巧明(1963-),男,教授,博士生导师,主要研究领域为中文信息处理。
  • 基金资助:
    国家自然科学基金(61836007,61472265,61772354)

Event Coreference Resolution Method Based on Attention Mechanism

CHENG Hao-yi, LI Pei-feng, ZHU Qiao-ming   

  1. (School of Computer Science and Technology,Soochow University,Suzhou,Jiangsu 215006,China);
    (Provincial Key Laboratory for Computer Information Processing Technology,Suzhou,Jiangsu 215006,China)
  • Received:2018-08-24 Online:2019-09-15 Published:2019-09-02

摘要: 事件同指消解是一项具有挑战性的自然语言处理任务,它在事件抽取、问答系统、阅读理解中有着重要的作用。文中提出了一种基于全局和局部信息,并具有全局推理机制的可分解注意力神经网络模型DANGL(Decomposable Attention Neural network based on Global and Local information),用于文档级的事件同指消解。神经网络模型DANGL与过去大部分以概率模型和图模型为基础的传统方法之间存在很大的区别。DANGL首先使用Bi-LSTM和CNN分别获取每个事件句的全局信息和局部信息;然后使用可分解注意力网络获取每个事件句中相对重要的信息;最后使用文档级全局推理模型进一步优化同指链。在TAC-KBP语料库上的实验显示,DANGL使用了少量的特征,且平均性能优于目前最好的基准系统。

关键词: 可分解注意力网络, 全局和局部信息, 全局推理, 事件同指

Abstract: Event coreference resolution is a challenging NLP task.It plays an import role in event extraction,QA system and reading comprehension.This paper introduced a decomposable attention neural network model DANGL with global inference mechanism based on remote and local information to document-level event coreference resolution.The neural network model DANGL is quite different from most traditional methods based on probabilistic models and graph models in the past.DANGL first uses Bi-LSTM and CNN to capture both the remote information and the local information of each event mention.Then,it applies the decomposable attention network to capture relatively important information in event mention.Finally,it employs a document-level global inference mechanism to further optimize the coreference chains.Experimental results on TAC-KBP show that DANGL uses a few features and outperforms the state-of-the-art baseline.

Key words: Decomposable attention network, Event coreference, Global inference, Remote and local information

中图分类号: 

  • TP391.1
[1]MOHAMMED M,MARWA A.Efficient Topic Detection System for Online Arabic News[OL].http://xueshu.baidu.com/usercenter/paper/show?paperid=ea8bdd0d677e034c-c43e7099b2e14b91&site=xueshu_se.
[2]CHENG L,GAO H K,WANG H B.A News Event Extraction Method in Chinese and Thai Languages Based on Dependency Tree Elements Combined with Rules[J].Software Guide,2018,17(7):49-56.
[3]SWAYAMDIPTA S,PARIKH A P,KWIATKOWSKI T.Multi-Mention Learning for Reading Comprehension with Neural Cascades[C]//Proceedings of theInternational Conference of Learning Representation.Vancouver:ICLR Press,2018.
[4]CHEN Z,JI H.Graph-based event coreference resolution[C]//Proceedings of the Association for Computational Linguistics.Singapore:ACL Press,2009:54-57.
[5]HAGHIGHI A,DAN K.Simple Coreference Resolution withRich Syntactic and Semantic Features[C]//Proceedings of the Empirical Methods in Natural Language Processing.Singapore:EMNLP Press,2009:1152-1161.
[6]BEJAN C A,HARABAGIU S.Unsupervised event coreference resolution with rich linguistic features[C]//Proceedings of the Association for Computational Linguistics.Sweden:ACL Press,2010:1412-1422.
[7] YANG B,CARDIE C,FRAZIER P.A Hierarchical Distance-dependent Bayesian Model for Event CoreferenceResolution[OL].https://arxiv.org/abs/1504.05929.
[8]CHEN Z,JI H.Graph-based event coreference resolution[C]//Proceedings of the Association for Computational Linguistics.Singapore:ACL Press,2009:54-57.
[9]CHEN Z,JI H,HARALICK R.A pairwise event coreferencemodel,feature impact and evaluation for event coreference resolution[C]//Proceedings of the Association for Computational Linguistics.Singapore:ACL Press,2009:17-22.
[10]LIU Z,ARAKI J,HOVY E,et al.Supervised within-document event coreference using information propagation[OL].http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.468.8992.
[11]LU J,NG V.Joint Learning for Event Coreference Resolution[C]//Proceedings of the Association for Computational Linguistics.Vancouver:ACL Press,2017:90-101.
[12]LIU Z,MITAMURA T,HOVY E.Graph-Based Decoding for Event Sequencing and Coreference Resolution[C]//Proceedings of the Computational Natural Language Learning.Belgium:CoNLL Press,2018.
[13]KRAUSE S,XU F,USZKOREIT H,et al.Event Linking with Sentential Features from Convolutional Neural Networks[C]//Proceedings of the Computational Natural Language Learning.Berlin:CoNLL Press,2016:239-249.
[14]CHOUBEY P K,HUANG R.Event Coreference Resolution by Iteratively Unfolding Inter-dependencies among Events[C]//Proceedings of the Empirical Methods in Natural Language Processing.Copenhagen:EMNLP Press,2017:2124-2133.
[15]PARIKH A P,TÄCKSTRÖM O,DAS D,et al.A Decomposable Attention Model for Natural Language Inference[C]//Procee-dings of the Empirical Methods in Natural Language Processing.Texas:EMNLP Press,2016:2249-2255.
[16]MITAMURA T,LIU Z,HOVY E.Overview of TAC-KBP 2015 Event Nugget Track[C]//Proceedings of the Tunnelling Association of Canada.Canada:TAC Press,2015.
[17]BAGGA A,BALDWIN B.Algorithms for Scoring Coreference Chains[C]//LREC Workshop on Linguistic Coreference.Spain:LREC Press,1998:563-566.
[18]LUO X.On coreference resolution performance metrics[C]//Proceedings of the Empirical Methods in Natural Language Processing.Vancouver:EMNLP Press,2005:25-32.
[19]VILAIN M,BURGER J,ABERDEEN J,et al.A Model-Theoretic Coreference Scoring Scheme[C]//Proceedings of the Association for Computational Linguistics.Cambridge:ACL Press,1995:45-52.
[20]LUO X.On coreference resolution performance metrics[C]//Proceedings of the Empirical Methods in Natural Language Processing.Vancouver:EMNLP Press,2005:25-32.
[1] 方杰, 李培峰, 朱巧明.
基于多注意力机制的事件同指消解方法
Employing Multi-attention Mechanism to Resolve Event Coreference
计算机科学, 2019, 46(8): 277-281. https://doi.org/10.11896/j.issn.1002-137X.2019.08.046
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!