End-to-end Neural Coreference Resolution
Kenton Leea b, Luheng Hea c, Mike Lewisd e, Luke Zettlemoyerb f
- aAllen Institute for Artificial Intelligence , Seattle WA
- bUniversity of Washington, Seattle, United States
- cGoogle (United States), Mountain View, United States
- dFacebook AI Research , Menlo Park , CA
- eUniversity of Pittsburgh, Pittsburgh, United States
- fSchool of Computer Science & Engineering , Univ. of Washington , Seattle , WA
Papers open inside SciRah from the original publisher or repository; SciRah never hosts them.
Abstract
We introduce the first end-to-end coreference resolution model and show that it significantly outperforms all previous work without using a syntactic parser or handengineered mention detector.The key idea is to directly consider all spans in a document as potential mentions and learn distributions over possible antecedents for each.The model computes span embeddings that combine context-dependent boundary representations with a headfinding attention mechanism.It is trained to maximize the marginal likelihood of gold antecedent spans from coreference clusters and is factored to enable aggressive pruning of potential mentions.Experiments demonstrate state-of-the-art performance, with a gain of 1.5 F1 on the OntoNotes benchmark and by 3.1 F1 using a 5-model ensemble, despite the fact that this is the first approach to be successfully trained with no external resources.
TL;DRThis work introduces the first end-to-end coreference resolution model, trained to maximize the marginal likelihood of gold antecedent spans from coreference clusters and is factored to enable aggressive pruning of potential mentions. (Semantic Scholar)