Using Document Level Cross-Event Inference to Improve Event Extraction

Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 789–797, Uppsala, Sweden, 11-16 July 2010. c⃝2010 Association for Computational Linguistics Using Document Level Cross-Event Inference to Improve Event Extraction Shasha Liao New York University 715 Broadway, 7th floor New York, NY 10003 USA liaoss@cs.nyu.edu Ralph Grishman New York University 715 Broadway, 7th floor New York, NY 10003 USA grishman@cs.nyu.edu Abstract Event extraction is a particularly challenging type of information extraction (IE). Most current event extraction systems rely on local information at the phrase or sentence level. However, this local context may be insufficient to resolve ambiguities in identifying particular types of events; information from a wider scope can serve to resolve some of these ambiguities. In this paper, we use document level information to improve the performance of ACE event extraction. In contrast to previous work, we do not limit ourselves to information about events of the same type, but rather use information about other types of events to make predictions or resolve ambiguities regarding a given event. We learn such relationships from the training corpus and use them to help predict the occurrence of events and event arguments in a text. Experiments show that we can get 9.0% (absolute) gain in trigger (event) classification, and more than 8% gain for argument (role) classification in ACE event extraction. 1 Introduction The goal of event extraction is to identify instances of a class of events in text. The ACE 2005 event extraction task involved a set of 33 generic event types and subtypes appearing frequently in the news. In addition to identifying the event itself, it also identifies all of the participants and attributes of each event; these are the entities that are involved in that event. Identifying an event and its participants and attributes is quite difficult because a larger field of view is often needed to understand how facts tie together. Sometimes it is difficult even for people to classify events from isolated sentences. From the sentence: (1) He left the company. it is hard to tell whether it is a Transport event in ACE, which means that he left the place; or an End-Position event, which means that he retired from the company. However, if we read the whole document, a clue like “he planned to go shopping before he went home” would give us confidence to tag it as a Transport event, while a clue like “They held a party for his retirement” would lead us to tag it as an End-Position event. Such clues are evidence from the same event type. However, sometimes another event type is also a good predictor. For example, if we find a Start-Position event like “he was named president three years ago”, we are also confident to tag (1) as End-Position event. Event argument identification also shares this benefit. Consider the following two sentences: (2) A bomb exploded in Bagdad; seven people died while 11 were injured. (3) A bomb exploded in Bagdad; the suspect got caught …

Content truncated to 500 words