A Rule-Based System for Unrestricted Bridging Resolution: Recognizing Bridging Anaphora and Finding Links to Antecedents

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 2082–2093, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics A Rule-Based System for Unrestricted Bridging Resolution: Recognizing Bridging Anaphora and Finding Links to Antecedents Yufang Hou1, Katja Markert2, Michael Strube1 1 Heidelberg Institute for Theoretical Studies gGmbH, Heidelberg, Germany (yufang.hou|michael.strube)@h-its.org 2School of Computing, University of Leeds, UK scskm@leeds.ac.uk Abstract Bridging resolution plays an important role in establishing (local) entity coher- ence. This paper proposes a rule-based approach for the challenging task of unre- stricted bridging resolution, where bridg- ing anaphors are not limited to defi- nite NPs and semantic relations between anaphors and their antecedents are not re- stricted to meronymic relations. The sys- tem consists of eight rules which target different relations based on linguistic in- sights. Our rule-based system significantly outperforms a reimplementation of a pre- vious rule-based system (Vieira and Poe- sio, 2000). Furthermore, it performs better than a learning-based approach which has access to the same knowledge resources as the rule-based system. Additionally, incorporating the rules and more features into the learning-based system yields a mi- nor improvement over the rule-based sys- tem. 1 Introduction Bridging resolution recovers the various non- identity relations between anaphora and an- tecedents. It plays an important role in establish- ing entity coherence in a text. In Example 1, the links between the bridging anaphors (The five as- tronauts and touchdown) and the antecedent (The space shuttle Atlantis) establish (local) entity co- herence.1 (1) The space shuttle Atlantis landed at a desert air strip at Edwards Air Force Base, Calif., ending a five-day mission that dispatched the Jupiter-bound Galileo space probe. The 1Examples are from OntoNotes (Weischedel et al., 2011). Bridging anaphora are typed in boldface; antecedents in ital- ics. five astronauts returned to Earth about three hours early because high winds had been pre- dicted at the landing site. Fog shrouded the base before touchdown. Bridging or associative anaphora has been widely discussed in the linguistic literature (Clark, 1975; Prince, 1981; Gundel et al., 1993; L¨obner, 1998). Poesio and Vieira (1998) and Bunescu (2003) include cases where antecedent and anaphor are coreferent but do not share the same head noun (different-head coreference). We follow our previous work (Hou et al., 2013b) and restrict bridging to non-coreferential cases. We also exclude comparative anaphora (Modjeska et al., 2003). Bridging resolution includes two subtasks: (1) recognizing bridging anaphors and (2) finding the correct antecedent among candidates. In recent empirical work, these two subtasks have been tackled separately: (Markert et al., 2012; Cahill and Riester, 2012; Rahman and Ng, 2012; Hou et al., 2013a) handle bridging recognition as part of information status (IS) classification, while (Poe- sio et al., 1997; Poesio et al., 2004; Markert et al., 2003; Lassalle and Denis, 2011; Hou et al., 2013b) concentrate on antecedent selection only, assuming that bridging recognition has already been performed. One exception is Vieira and Poe- sio (2000). They propose a rule-based system for processing definite NPs. However, they …

Content truncated to 500 words