EM Works for Pronoun Anaphora Resolution

Proceedings of the 12th Conference of the European Chapter of the ACL, pages 148–156, Athens, Greece, 30 March – 3 April 2009. c⃝2009 Association for Computational Linguistics EM Works for Pronoun Anaphora Resolution Eugene Charniak and Micha Elsner Brown Laboratory for Linguistic Information Processing (BLLIP) Brown University Providence, RI 02912 {ec,melsner}@cs.brown.edu Abstract We present an algorithm for pronoun- anaphora (in English) that uses Expecta- tion Maximization (EM) to learn virtually all of its parameters in an unsupervised fashion. While EM frequently fails to find good models for the tasks to which it is set, in this case it works quite well. We have compared it to several systems avail- able on the web (all we have found so far). Our program significantly outperforms all of them. The algorithm is fast and robust, and has been made publically available for downloading. 1 Introduction We present a new system for resolving (per- sonal) pronoun anaphora1. We believe it is of interest for two reasons. First, virtually all of its parameters are learned via the expectation- maximization algorithm (EM). While EM has worked quite well for a few tasks, notably ma- chine translations (starting with the IBM models 1-5 (Brown et al., 1993), it has not had success in most others, such as part-of-speech tagging (Meri- aldo, 1991), named-entity recognition (Collins and Singer, 1999) and context-free-grammar in- duction (numerous attempts, too many to men- tion). Thus understanding the abilities and limi- tations of EM is very much a topic of interest. We present this work as a positive data-point in this ongoing discussion. Secondly, and perhaps more importantly, is the system’s performance. Remarkably, there are very few systems for actually doing pronoun anaphora available on the web. By emailing the corpora- list the other members of the list pointed us to 1The system, the Ge corpus, and the model described here can be downloaded from http://bllip.cs.brown.edu/download/emPronoun.tar.gz. four. We present a head to head evaluation and find that our performance is significantly better than the competition. 2 Previous Work The literature on pronominal anaphora is quite large, and we cannot hope to do justice to it here. Rather we limit ourselves to particular papers and systems that have had the greatest impact on, and similarity to, ours. Probably the closest approach to our own is Cherry and Bergsma (2005), which also presents an EM approach to pronoun resolution, and ob- tains quite successful results. Our work improves upon theirs in several dimensions. Firstly, they do not distinguish antecedents of non-reflexive pronouns based on syntax (for instance, subjects and objects). Both previous work (cf. Tetreault (2001) discussed below) and our present results find these distinctions extremely helpful. Sec- ondly, their system relies on a separate prepro- cessing stage to classify non-anaphoric pronouns, and mark the gender of certain NPs (Mr., Mrs. and some first names). This allows the incorpo- ration of external data and learning systems, but conversely, it requires these decisions to be made sequentially. Our system classifies non-anaphoric pronouns jointly, and learns gender …

Content truncated to 500 words