A neighborhood relevance model for entity linking

Entity Linking is the task of mapping mentions in documents to entities in a knowledge base. One of the crucial tasks is to identify the disambiguating context of the mention, and joint assignment models leverage the relationships within the knowledge base. We demonstrate how joint assignment models can be approximated with information retrieval. We build on pseudo-relevance feedback and use the source corpus to build an neighborhood relevance model that we show is more effective than local models for ranking KB entities. Our results demonstrate that simple text based features combined with a supervised Learning to Rank model result a model that matches or outperforms the top performing system on in-KB accuracy in the TAC KBP entity linking task.