Employing Compositional Semantics and Discourse Consistency in Chinese Event Extraction

Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 1006–1016, Jeju Island, Korea, 12–14 July 2012. c⃝2012 Association for Computational Linguistics Employing Compositional Semantics and Discourse Consistency in Chinese Event Extraction Peifeng Li, Guodong Zhou, Qiaoming Zhu, Libin Hou School of Computer Science & Technology Soochow University, Suzhou, 215006, China {pfli, gdzhou, qmzhu, 20094227021}@suda.edu.cn Abstract Current Chinese event extraction systems suffer much from two problems in trigger identification: unknown triggers and word segmentation errors to known triggers. To resolve these problems, this paper proposes two novel inference mechanisms to explore special characteristics in Chinese via compositional semantics inside Chinese triggers and discourse consistency between Chinese trigger mentions. Evaluation on the ACE 2005 Chinese corpus justifies the effectiveness of our approach over a strong baseline. 1 Introduction Event extraction, a classic information extraction task, is to identify instances of a predefined event type and can be typically divided into four subtasks: trigger identification, trigger type determination, argument identification and argument role determination. In the literature, most studies focus on English event extraction and have achieved certain success (e.g. Grishman et al., 2005; Ahn, 2006; Hardy et al., 2006; Maslennikov and Chua, 2007; Finkel et al., 2005; Ji and Grishman, 2008; Patwardhan and Riloff, 2009, 2011; Liao and Grishman 2010; Hong et al., 2011). In comparison, there are few successful stories regarding Chinese event extraction due to special characteristics in Chinese trigger identification. In particular, there are two major reasons for the low performance: unknown triggers 1 and word segmentation errors to known triggers. Table 1 gives the statistics of unknown triggers and word segmentation errors to known triggers in both the 1 In this paper, a trigger word/phrase occurring in the training data is called a known trigger and otherwise, an unknown trigger. ACE 2005 Chinese and English corpora2 using 10- fold cross-validation. In each validation, we leave 10% trigger mentions as the test set and the remaining ones as the training set. If a mention in the test set doesn’t occurred in the training set, we regard it as an unknown trigger. It shows that these two cases cover almost 30% of Chinese trigger mentions while this figure reduces to only about 9% in English. It also shows that given the same number of event mentions, there are 30% more different triggers in Chinese than that in English. This justifies the low performance (specifically, the recall) of a Chinese event extraction system, which normally extracts those known triggers occurring in the training data as candidate instances and uses a classifier to distinguish correct triggers from wrong ones. Language Chinese English %unknown triggers 33.7% 18.5% %unknown trigger mentions 20.9% 8.9% %word segmentation errors to known trigger mentions 8.7% 0% #triggers 763 586 Table 1. Statistics: a comparison between Chinese and English event extraction with regard to unknown triggers and word segmentation errors to known triggers. Note that word segmentation only applies to Chinese. In this paper, we propose two novel inference …

Content truncated to 500 words