They Exist! Introducing Plural Mentions to Coreference Resolution and Entity Linking

Proceedings of the 27th International Conference on Computational Linguistics, pages 24–34 Santa Fe, New Mexico, USA, August 20-26, 2018. 24 They Exist! Introducing Plural Mentions to Coreference Resolution and Entity Linking Ethan Zhou Computer Science Emory University Atlanta, GA 30322 ethan.zhou@emory.edu Jinho D. Choi Computer Science Emory University Atlanta, GA 30322 jinho.choi@emory.edu Abstract This paper analyzes arguably the most challenging yet under-explored aspect of resolution tasks such as coreference resolution and entity linking, that is the resolution of plural mentions. Unlike singular mentions each of which represents one entity, plural mentions stand for multiple entities. To tackle this aspect, we take the character identification corpus from the SemEval 2018 shared task that consists of entity annotation for singular mentions, and expand it by adding annotation for plural mentions. We then introduce a novel coreference resolution algorithm that selectively creates clusters to handle both singular and plural mentions, and also a deep learning-based entity linking model that jointly handles both types of mentions through multi-task learning. Adjusted evaluation metrics are proposed for these tasks as well to handle the uniqueness of plural mentions. Our experiments show that the new coreference resolution and entity linking models significantly outperform traditional models designed only for singular mentions. To the best of our knowledge, this is the first time that plural mentions are thoroughly analyzed for these two resolution tasks. 1 Introduction Resolution tasks such as coreference resolution and entity linking are challenging because they require a holistic view of a document (or across multiple documents) to find correct entities. Although many models have been proposed for these tasks (Clark and Manning, 2016; Francis-Landau et al., 2016; Wiseman et al., 2016; Gupta et al., 2017; Lee et al., 2017), most of them are focused on singular mentions such that they are insufficient for resolving the other type of mentions, plural, although the amount of plural mentions is not negligible in practice.1 Table 1 illustrates how mentions are annotated for coreference resolution by the CoNLL’12 shared task (Pradhan et al., 2012) and our proposed work. In the CoNLL’12 annotation, the plural mention They8 is grouped with the noun phrase [Mary1 and John2]3; however, the other plural mention We7 becomes a singleton because there is no noun phrase representing such an entity. Since CoNLL’12 limits each plural mention to be linked to a single noun phrase, it loses connections to individual entities that exist within the document but not grouped as a noun phrase. Document [Mary1 and John2]3 came to see me4 yesterday. She5 looked happy, and so did he6. We7 had a great time together. They8 left around noon. CoNLL’12 {Mary1, She5}, {John2, he6}, {[Mary1 and John2]3, They8}, {me4}, {We7} Our Work {Mary1, She5, We7, They8}, {John2, he6, We7, They8}, {me4, We7} Table 1: Snippets of how mentions are annotated by the CoNLL’12 shared task and our work. In our work, the plural mentions We7 and They8 are linked to multiple entities that those mentions refer to. This allows higher-level NLP tasks such as question answering …

Content truncated to 500 words