A Tree Kernel-Based Unified Framework for Chinese Zero Anaphora Resolution

Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 882–891, MIT, Massachusetts, USA, 9-11 October 2010. c⃝2010 Association for Computational Linguistics A Tree Kernel-based Unified Framework for Chinese Zero Anaphora Resolution Fang Kong Guodong Zhou* JiangSu Provincial Key Lab for Computer Information Processing Technology School of Computer Science and Technology Soochow University {kongfang, gdzhou}@suda.edu.cn * Corresponding author Abstract This paper proposes a unified framework for zero anaphora resolution, which can be di- vided into three sub-tasks: zero anaphor detec- tion, anaphoricity determination and antecedent identification. In particular, all the three sub-tasks are addressed using tree ker- nel-based methods with appropriate syntactic parse tree structures. Experimental results on a Chinese zero anaphora corpus show that the proposed tree kernel-based methods signifi- cantly outperform the feature-based ones. This indicates the critical role of the structural in- formation in zero anaphora resolution and the necessity of tree kernel-based methods in modeling such structural information. To our best knowledge, this is the first systematic work dealing with all the three sub-tasks in Chinese zero anaphora resolution via a unified framework. Moreover, we release a Chinese zero anaphora corpus of 100 documents, which adds a layer of annotation to the manu- ally-parsed sentences in the Chinese Treebank (CTB) 6.0. 1 Introduction As one of the most important techniques in dis- course analysis, anaphora resolution has been a focus of research in Natural Language Processing (NLP) for decades and achieved much success in English recently (e.g. Soon et al. 2001; Ng and Cardie 2002; Yang et al. 2003, 2008; Kong et al. 2009). However, there is little work on anaphora reso- lution in Chinese. A major reason for this phe- nomenon is that Chinese, unlike English, is a pro- drop language, whereas in English, definite noun phrases (e.g. the company) and overt pronouns (e.g. he) are frequently employed as referring expres- sions, which refer to preceding entities. Kim (2000) compared the use of overt subjects in English and Chinese. He found that overt subjects occupy over 96% in English, while this percentage drops to only 64% in Chinese. This indicates the prevalence of zero anaphors in Chinese and the necessity of zero anaphora resolution in Chinese anaphora reso- lution. Since zero anaphors give little hints (e.g. number or gender) about their possible antecedents, zero anaphora resolution is much more challenging than traditional anaphora resolution. Although Chinese zero anaphora has been widely studied in the linguistics research (Li and Thompson 1979; Li 2004), only a small body of prior work in computational linguistics deals with Chinese zero anaphora resolution (Converse 2006; Zhao and Ng 2007). Moreover, zero anaphor de- tection, as a critical component for real applica- tions of zero anaphora resolution, has been largely ignored. This paper proposes a unified framework for Chinese zero anaphora resolution, which can be divided into three sub-tasks: zero anaphor detec- tion, which detects zero anaphors from a text, ana- phoricity determination, which determines whether a zero anaphor is anaphoric or not, and antecedent identification, which finds …

Content truncated to 500 words