Approaches to biomedical coreference resolution

Coreference resolution is an important task in natural language processing which aims to group the mention pairs referring to a single entity. In the biomedical domain, it significantly poses some unique challenges. In this work, we make use of both hand-crafted features and neural word embedding based features to solve the task of coreference resolution on a standard benchmark biomedical coreference dataset, i.e the BioNLP-2011 Protein Coreference data. Experimental results show that the neural model performs significantly better in terms of mention-referent linking when compared to the hand-crafted feature-based coreference resolution approaches.