A Unified Event Coreference Resolution by Integrating Multiple Resolvers

12471

Authors

nan

Abstract

Event coreference is an important and compli- cated task in cascaded event template extraction and other natural language processing tasks. De- spite its importance, it was merely discussed in previous studies. In this paper, we present a glo- bally optimized coreference resolution system dedicated to various sophisticated event corefe- rence phenomena. Seven resolvers for both event and object coreference cases are utilized, which include three new resolvers for event coreference resolution. Three enhancements are further pro- posed at both mention pair detection and chain formation levels. First, the object coreference re- solvers are used to effectively reduce the false positive cases for event coreference. Second, A revised instance selection scheme is proposed to improve link level mention-pair model perfor- mances. Last but not least, an efficient and glo- bally optimized graph partitioning model is em- ployed for coreference chain formation using spectral partitioning which allows the incorpora- tion of pronoun coreference information. The three techniques contribute to a significant im- provement of 8.54% in B3 F-score for event co- reference resolution on OntoNotes 2.0 corpus. 1

Keywords

nan

Full Content

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 102–110, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP A Unified Event Coreference Resolution by Integrating Multiple Resolvers Bin Chen1, Jian Su2, Sinno Jialin Pan3 and Chew Lim Tan4 1,2,3Institute for Infocomm Research, Singapore 4National University of Singapore 1bchen,2sujian,3jspan@i2r.a-star.edu.sg 4tancl@comp.nus.edu.sg Abstract Event coreference is an important and compli- cated task in cascaded event template extraction and other natural language processing tasks. De- spite its importance, it was merely discussed in previous studies. In this paper, we present a glo- bally optimized coreference resolution system dedicated to various sophisticated event corefe- rence phenomena. Seven resolvers for both event and object coreference cases are utilized, which include three new resolvers for event coreference resolution. Three enhancements are further pro- posed at both mention pair detection and chain formation levels. First, the object coreference re- solvers are used to effectively reduce the false positive cases for event coreference. Second, A revised instance selection scheme is proposed to improve link level mention-pair model perfor- mances. Last but not least, an efficient and glo- bally optimized graph partitioning model is em- ployed for coreference chain formation using spectral partitioning which allows the incorpora- tion of pronoun coreference information. The three techniques contribute to a significant im- provement of 8.54% in B3 F-score for event co- reference resolution on OntoNotes 2.0 corpus. 1 Introduction Coreference resolution, the task of resolving and linking different mentions of the same ob- ject/event in a text, is important for an intelligent text processing system. The resolved coreferent mentions form a coreference chain representing a particular object/event. Following the natural order in the texts, any two consecutive mentions in a coreference chain form an anaphoric pair with the latter mention referring back to the prior one. The latter mention is called the anaphor while the prior one is named as the antecedent. Most of previous works on coreference resolu- tion such as (Soon et al, 2001; Yang et al, 2006), aimed at object coreference which both the ana- phor and its antecedent are mentions of the same real world object such as person, location and organization. In contrast, an event coreference as defined in (Asher, 1993) is an anaphoric refer- ence to an event, fact, and proposition which is representative of eventuality and abstract entities. In the following example: “Israel has [fired] missiles on the offices of the Palestinian Authority. [It] has caused 7 deaths with many injuries… Israel helicopter gunships [fired] across the Gaza Strip for more than two hours. [The attack] in Gaza has been said to cause more violence in Gaza and West Bank and terminate the current round of mid-East peace talk in an unexpected way.” The four mentions here, [fired], [it], [fired] and [the attack] are referring to the same event (an Israel attack in Gaza Strip on Palestinian Author- ity). The pronouns noun phrases and action verbs are taken as the representation of events which is also in line with OntoNotes …

Content truncated to 500 words

Labeled Text

Event coreference is an important and complicated task in cascaded event template extraction and other natural language processing tasks. Despite its importance, it was merely discussed in previous studies. In this paper, we present a globally optimized coreference resolution system dedicated to various sophisticated event coreference phenomena. Seven resolvers for both event and object coreference cases are utilized, which include three new resolvers for event coreference resolution. Three enhancements are further proposed at both mention pair detection and chain formation levels. First, the object coreference resolvers are used to effectively reduce the false positive cases for event coreference. Second, A revised instance selection scheme is proposed to improve link level mention-pair model performances. Last but not least, an efficient and globally optimized graph partitioning model is employed for coreference chain formation using spectral partitioning which allows the incorporation of pronoun coreference information. The three techniques contribute to a significant improvement of 8.54% in B3 F-score for event coreference resolution on OntoNotes 2.0 corpus. 1

Publication Details

Paper ID:
12471
Year:
2011
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan