Other-Anaphora Resolution in Biomedical Texts with Automatically Mined Patterns

Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008), pages 121–128 Manchester, August 2008 Other-Anaphora Resolution in Biomedical Texts with Automatically Mined Patterns Chen Bin#, Yang Xiaofeng$, Su Jian^ and Tan Chew Lim* #*School of Computing, National University of Singapore $^Institute for Infocomm Research, A-STAR, Singapore {#chenbin, *tancl}@comp.nus.edu.sg {$xiaofengy, ^sujian}@i2r.a-star.edu.sg  Abstract This paper proposes an other-anaphora resolution approach in bio-medical texts. It utilizes automatically mined patterns to discover the semantic relation between an anaphor and a candidate antecedent. The knowledge from lexical patterns is incor- porated in a machine learning framework to perform anaphora resolution. The ex- periments show that machine learning approach combined with the auto-mined knowledge is effective for other- anaphora resolution in the biomedical domain. Our system with auto-mined pat- terns gives an accuracy of 56.5%., yield- ing 16.2% improvement against the base- line system without pattern features, and 9% improvement against the system us- ing manually designed patterns. 1 Introduction The last decade has seen an explosive growth in the amount of textual information in biomedi- cine. There is a need for an effective and effi- cient text-mining system to gather and utilize the knowledge encoded in the biomedical literature. For a correct discourse analysis, a text-mining system should have the capability of understand- ing the reference relations among different ex- pressions in texts. Hence, anaphor resolution, the task of resolving a given text expression to its referred expression in prior texts, is important for an intelligent text processing system. © 2008. Licensed under the Creative Commons Attri- bution-Noncommercial-Share Alike 3.0 Unported license (http://creativecommons.org/licenses/by-nc- sa/3.0/). Some rights reserved. In linguistics, an expression that points back to a previously mentioned expression is called an anaphor, and the expression being referred to by the anaphor is called its antecedent. Most pre- vious work on anaphora resolution aims at identi- ty-anaphora in which both an anaphor and its antecedent are mentions of the same entity. In this paper, we focus on a special type of anaphora resolution, namely, other-anaphora resolution, in which an anaphor to be resolved has a prefix modifier “other” or “another”. The antecedent of an other-anaphor is a complement expression to the anaphor in a super set. In other words, an other-anaphor is a set of elements ex- cluding the element(s) specified by the antece- dent. If the modifier “other” or “another” is re- moved, an anaphor becomes the super set includ- ing the antecedent. Thus, other-anaphora in fact represents a “part-whole” relation. Consider the following text “IL-10 inhibits nuclear stimulation of nuclear factor kappa B (NF kappa B). Several other transcription factors including NF- IL-6, AP-1, AP-2, GR, CREB, Oct-1, and Sp-1 are not affected by IL-10.” Here, the expression “other transcription fac- tors” is an other-anaphor, while the “NF kappa B” is its antecedent. The anaphor refers to any transcription factors except the antecedent. By removing the lexical modifier “other”, we can get a supper set “transcription factors” that in- cludes the antecedent. The anaphor and antece- dent thus have a …

Content truncated to 500 words