The Life and Death of Discourse Entities: Identifying Singleton Mentions

A discourse typically involves numerous entities, but few are mentioned more than once. Distinguishing discourse entities that die out after just one mention (singletons) from those that lead longer lives (coreferent) would benefit NLP applications such as coreference resolution, protagonist identification, topic modeling, and discourse coherence. We build a logistic regression model for predicting the singleton/coreferent distinction, drawing on linguistic insights about how discourse entity lifespans are affected by syntactic and semantic features. The model is effective in its own right (78% accuracy), and incorporating it into a state-of-the-art coreference resolution system yields a significant improvement.