Unrestricted Bridging Resolution

Abstract

In contrast to identity anaphors which indicate coreference between a noun phrase and its antecedent, bridging anaphors link to their antecedent(s) via lexico-semantic, frame or encyclopedic relations which do not lead to coreference. Bridging resolution involves recognizing bridging anaphora and finding links to antecedents. In contrast to most prior work, we tackle both problems in this article. Our work is - to the best of our knowledge - also the first bridging
resolution system that does not impose any restrictions on the type of bridging anaphora or relations between anaphor and antecedent.

We create a corpus (ISNotes) where bridging is reliably annotated. We break new ground by considering all relations and anaphora/antecedent types and show that the variety of bridging anaphora is much higher than reported previously. We then solve the problem of bridging resolution using a two-stage statistical global inference method. Given all mentions in a document, the first stage, bridging anaphora recognition, recognizes bridging anaphors as a subtask of learning fine-grained information status (IS). Each mention in a text gets assigned one IS class, bridging being one possible class. We use a cascading collective classification method where (i) collective classification allows us to investigate relations among several mentions and autocorrelation among IS classes and (ii) cascaded classification allows us to tackle class imbalance, important for minority classes such as bridging. The second stage, bridging antecedent selection,
finds the antecedents for all predicted bridging anaphors. Semantically or syntactically related anaphors in a document are likely to share the same antecedent. We use this phenomenon, which we call sibling anaphors, in a joint inference model. Both bridging recognition as well as bridging antecedent selection are feature-based and discriminatively trained on ISNotes. Therefore, apart
from the novel use of joint inference for these problems, we also make important contributions in feature design with regard to semantic and salience features.