LingMess: Linguistically Informed Multi Expert Scorers for Coreference Resolution
This work addresses coreference resolution for natural language processing by proposing a novel method to handle diverse linguistic challenges, representing an incremental improvement over single-scorer models.
The authors tackled the problem of coreference resolution by introducing LingMess, a model that uses multiple pairwise scorers for different linguistic categories, which substantially improves pairwise scores for most categories and outperforms cluster-level performance on Ontonotes and five additional datasets.
While coreference resolution typically involves various linguistic challenges, recent models are based on a single pairwise scorer for all types of pairs. We present LingMess, a new coreference model that defines different categories of coreference cases and optimize multiple pairwise scorers, where each scorer learns a specific set of linguistic challenges. Our model substantially improves pairwise scores for most categories and outperforms cluster-level performance on Ontonotes and 5 additional datasets. Our model is available in https://github.com/shon-otmazgin/lingmess-coref