Machine Learning for Coreference Resolution: From Local Classification to Global Ranking

Proceedings of the 43rd Annual Meeting of the ACL, pages 157–164, Ann Arbor, June 2005. c⃝2005 Association for Computational Linguistics Machine Learning for Coreference Resolution: From Local Classification to Global Ranking Vincent Ng Human Language Technology Research Institute University of Texas at Dallas Richardson, TX 75083-0688 vince@hlt.utdallas.edu Abstract In this paper, we view coreference reso- lution as a problem of ranking candidate partitions generated by different coref- erence systems. We propose a set of partition-based features to learn a rank- ing model for distinguishing good and bad partitions. Our approach compares fa- vorably to two state-of-the-art coreference systems when evaluated on three standard coreference data sets. 1 Introduction Recent research in coreference resolution — the problem of determining which noun phrases (NPs) in a text or dialogue refer to which real-world entity — has exhibited a shift from knowledge- based approaches to data-driven approaches, yield- ing learning-based coreference systems that rival their hand-crafted counterparts in performance (e.g., Soon et al. (2001), Ng and Cardie (2002b), Strube et al. (2002), Yang et al. (2003), Luo et al. (2004)). The central idea behind the majority of these learning- based approaches is to recast coreference resolution as a binary classification task. Specifically, a clas- sifier is first trained to determine whether two NPs in a document are co-referring or not. A separate clustering mechanism then coordinates the possibly contradictory pairwise coreference classification de- cisions and constructs a partition on the given set of NPs, with one cluster for each set of coreferent NPs. Though reasonably successful, this “standard” ap- proach is not as robust as one may think. First, de- sign decisions such as the choice of the learning al- gorithm and the clustering procedure are apparently critical to system performance, but are often made in an ad-hoc and unprincipled manner that may be suboptimal from an empirical point of view. Second, this approach makes no attempt to search through the space of possible partitions when given a set of NPs to be clustered, employing instead a greedy clustering procedure to construct a partition that may be far from optimal. Another potential weakness of this approach con- cerns its inability to directly optimize for clustering- level accuracy: the coreference classifier is trained and optimized independently of the clustering pro- cedure to be used, and hence improvements in clas- sification accuracy do not guarantee corresponding improvements in clustering-level accuracy. Our goal in this paper is to improve the robustness of the standard approach by addressing the above weaknesses. Specifically, we propose the following procedure for coreference resolution: given a set of NPs to be clustered, (1) use � pre-selected learning- based coreference systems to generate � candidate partitions of the NPs, and then (2) apply an auto- matically acquired ranking model to rank these can- didate hypotheses, selecting the best one to be the fi- nal partition. The key features of this approach are: Minimal human decision making. In contrast to the standard approach, our method obviates, to a large extent, the need to …

Content truncated to 500 words