Round Corner
Department of Computer and Information Science


Entity linking in a conversational setting

Entity linking is the task of recognizing mentions of specific entities in text (e.g., persons, locations, products) and linking these mentions to the corresponding entries in an underlying knowledge repository (e.g., Wikipedia). See this book chapter for an explanation of the problem and the Google Cloud Natural Language API for an entity linking service.

It remains an open question how well existing techniques work in a conversational setting (think SIRI or the Google Assistant), where user utterances are short and there may not be enough context for full disambiguation (i.e., the system would need to ask a clarification question). This project aims at developing an entity linking approach for conversational search. Specific objectives include (i) evaluating how well existing entity linking techniques work out-of-the-box, (ii) leveraging context (previous utterances) to improve performance, and (iii) asking clarification questions when there is not enough context for disambiguation.



Krisztian Balog Krisztian Balog
Adjunct Professor
248 IT-bygget
NTNU logo