Exploring Local and Global Semantic Information for Event Pronoun Resolution

Proceedings of COLING 2012: Technical Papers, pages 1475–1488, COLING 2012, Mumbai, December 2012. *Corresponding author Exploring Local and Global Semantic Information for Event Pronoun Resolution Fang Kong1,2 Guodong Zhou1* (1) School of Computer Science and Technology, Soochow University of China, 1 Shizi Street, Suzhou, China 215006 (2) Department of Computer Science, National University of Singapore, 13 Computing Drive, Singapore 117417 kongfang@suda.edu.cn, gdzhou@suda.edu.cn ABSTRACT Event anaphora resolution plays a critical role in discourse analysis. This paper focuses on improving event pronoun resolution using both local and global semantic information. In particular, a predicate-argument structure is proposed to represent the local semantic information about an event while the global semantic information is represented by the entity coreference chains related with various arguments in the predicate-argument structure to complement its locality. Evaluation on the OntoNotes English corpus shows the effectiveness of local and global semantic information for event pronoun resolution. KEYWORDS: event pronoun, semantic information, corefernce resolution, predicate-argument structure 1475 1 Introduction As one of the most important techniques in discourse analysis, anaphora resolution aims to resolve a given mention to its referred expression in a text and has been a focus of research in Natural Language Processing (NLP) for decades. According to the nature of the referred expression, anaphora resolution can be categorized into entity anaphora resolution and event anaphora resolution. While most studies focus on entity anaphora resolution and have achieved much success recently (e.g. Soon et al. 2001; Ng and Cardie 2002; Ng 2007, 2009; Yang et al.2004, 2006, 2008; Kong et al. 2009,2010), there are only a few studies on event anaphora resolution (Byron, 2002; Pradhan et al., 2007; Chen et al. 2010a, 2010b; Kong and Zhou, 2011). In this paper, we address event pronoun resolution, the most difficult type of event anaphora resolution due to the least discriminative information that an event pronoun can provide. Here, an event pronoun is a pronoun whose antecedent refers to an event. In particular, we focus on improving event pronoun resolution using both local and global semantic information. For local semantic information, we employ a shallow semantic parser to extract the predicate-argument structure in a sentence to represent an involved event. In order to complement the locality of the predicate-argument structure, we consider the global semantic information via the entity coreference chains related with various arguments in the predicate-argument structure. The rest of this paper is organized as follows. Section 2 describes the event anaphora resolution task. Section 3 briefly introduces the related work on event anaphora resolution. Section 4 presents our baseline framework which combines various kinds of flat features and a structured parse tree for event pronoun resolution. Section 5 explores both local and global semantic information for event pronoun resolution. Section 6 reports the experimental results. Finally, we conclude our work in Section 7. 2 Task Description While entity anaphora focuses on mentions of an entity, event anaphora looks into mentions that refer to an event. In this paper, we consider event anaphora resolution. Consider the following examples: a) …

Content truncated to 500 words