Building Chinese Event Type Paradigm Based on Trigger Clustering

International Joint Conference on Natural Language Processing, pages 311–319, Nagoya, Japan, 14-18 October 2013. Building Chinese Event Type Paradigm Based on Trigger Clustering Xiao Ding, Bing Qin, Ting Liu* Research Center for Social Computing and Information Retrieval Harbin Institute of Technology, China {xding, qinb, tliu}@ir.hit.edu.cn Abstract Traditional Event Extraction mainly focuses on event type identification and event partici- pants extraction based on pre-specified event type annotations. However, different domains have different event type paradigms. When transferring to a new domain, we have to build a new event type paradigm. It is a costly task to discover and annotate event types manually. To address this problem, this paper proposes a novel approach of building an event type para- digm by clustering event triggers. Based on the trigger clusters, the event type paradigm can be built automatically. Experimental re- sults on three different corpora – ACE (small, homogeneous, open corpus), Financial News and Musical News (large scale, specific do- main, web corpus) indicate that our method can effectively build an event type paradigm and can be easily adapted to new domains. 1 Introduction Event extraction techniques have been widely used in several different specific domains, such as musical reports (Ding et al., 2011), financial analysis (Lee et al., 2003), biomedical investiga- tion (Yakushiji et al., 2001) and legal documents (Schilder et al., 2007). Traditional event extrac- tion systems achieved excellent performance in some important information extraction bench- marks, such as MUC (Message Understanding Conference, Chinechor et al., 1994) and ACE (Automatic Content Extraction). However, most of these methods require pre-specified event types as their prior knowledge. For example, ACE defines an event as a specific occurrence involving participants, and it annotates 8 types and 33 subtypes of events (LDC, 2005). Howev- er, building an event type paradigm in this way not only requires massive human effort but also tends to be very data dependent. As a result, it  * Email correspondence may prevent the event extraction from being widely applicable. Since event types among do- mains are different, the event type paradigm of ACE, which does not define music related events, is useless for the music domain event extraction. So we have to build a totally different event type paradigm for the music domain from scratch. Recently, some researchers have been aware of the limitations of only considering pre-defined paradigm as well. In the same vein, some studies work on the problem of relation extraction (Chambers and Jurasky, 2011 and 2009; Poon and Domingos, 2009 and 2008; Yates and Etzio- ni, 2009). Rosenfeld and Feldman (2006) built a high-performance unsupervised relation extrac- tion system without target relations in advance. Hasegawa et al. (2004) discovered relations among named entities from large corpora by clustering pairs of named entities. However, most of the above work focuses on relation ex- traction rather than event extraction. In contrast to the well-studied problem of rela- tion extraction, only a few works focused on event extraction. For example, Li (2010) pro- posed a domain-independent novel event discov- …

Content truncated to 500 words