分类器模型是目前识别因果关系的主要模型，该方法存在的问题是只考虑2个事件之间的关系，没有考虑同一文档中其他关联事件所包含的信息，识别结果往往存在逻辑矛盾。该文提出了一个中文事件因果关系识别的全局优化方法，该方法采用整数线性规划（integer linear programming，ILP）的推理方法，对基本逻辑关系、因果标志词、事件类型、论元信息进行有效约束，以文档为单位来优化因果关系识别。在该文标注语料上的实验结果表明：与分类器方法相比，该文提出的全局优化方法的F1值提升了5.54%。
Classifier-based models are widely used to identify causal relations between events. However, these models only consider the relationship between two specified events while ignoring related events. Thus, the results may have many logical contradictions. This paper presents a global optimization approach to recognize causal relations between events using an inference method based on integer linear programming (ILP). This approach introduces various kinds of constraints, i.e., a basic logical relationship, causal signal words, event types and argument information constraints to improve the performance. Tests on an annotated corpus show that this global optimization approach improves the F1 score by 5.54% compared with a classifier-based model.
[1] Wolff P. Representing causation[J]. Journal of Experimental Psychology:General, 2007, 136(1):82-111.[2] Mirza P, Sprugnoli R, Tonelli S, et al. Annotating causality in the TempEval-3 corpus[C]//Proceedings of EACL 2014 Workshop on Computational Approaches to Causality in Language. Gothenburg, Sweden:Association for Computational Linguistics, 2014:10-19.[3] Girju R. Automatic detection of causal relations for question answering[C]//Proceedings of the ACL 2003 Workshop on Multilingual Summarization and Question Answering. Sapporo, Japan:Association for Computational Linguistics, 2003:76-83.[4] Khoo C S G, Chan S, Niu Y. Extracting causal knowledge from a medical database using graphical patterns[C]//Proceedings of the 38th Annual Meeting on Association for Computational Linguistics. Hong Kong, China:Association for Computational Linguistics, 2000:336-343.[5] Ittoo A, Bouma G. Extracting Explicit and Implicit Causal Relations from Sparse, Domain-specific Texts[M]. Berlin Heidelberg:Springer, 2011.[6] Radinsky K, Davidovich S, Markovitch S. Learning causality for news events prediction[C]//International Conference on World Wide Web. Lyon, France:ACM, 2012:909-918.[7] Bethard S, Corvey W J, Klingenstein S, et al. Building a corpus of temporal-causal structure[C]//Proceedings of the 6th International Conference on Language Resources and Evaluation. Marrakech, Morocco:European Language Resources Association, 2008:908-915.[8] Rink B, Bejan C A, Harabagiu S M. Learning textual graph patterns to detect causal event relations[C]//The 23rd International Florida Artificial Intelligence Research Society Conference. Florida, FL, USA:Association for the Advancement of Artificial Intelligence, 2010:265-270.[9] Do Q X, Chan Y S, Roth D. Minimally supervised event causality identification[C]//Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing. Edinburgh, UK:Association for Computational Linguistics, 2011:294-303.[10] Chambers N, Jurafsky D. Jointly combining implicit constraints improves temporal ordering[C]//Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing. Honolulu, HI, USA:Association for Computational Linguistics, 2008:698-706.[11] 郑新. 中文事件时序关系识别与推理方法研究[D]. 苏州:苏州大学, 2015.ZHENG Xin. Research on Temporal Relation Recognition and Inference Between Chinese Events[D]. Suzhou:Soochow University, 2015. (in Chinese)[12] Denis P, Baldridge J. Joint determination of anaphoricity and coreference resolution using integer programming[C]//Proceedings of the Human Language Technologies:The Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL-HLT 2007). Rochester, NY, USA:Association for Computational Linguistics, 2007:236-243.[13] 黄一龙, 李培峰, 朱巧明. 中文事件相关性语料库构建及识别方法[J]. 计算机工程与科学, 2015, 37(12):2306-2311.HUANG Yilong, LI Peifeng, ZHU Qiaoming. The construction of Chinese relevant event corpus and its recognition approach[J]. Computer Engineering and Science, 2015, 37(12):2306-2311. (in Chinese)
