ENCORE. Unrestricted Cross-Document Event Coreference for Dutch

Start - End 
2020 - 2024 (ongoing)
Type 
Research Focus 
Research Period 
Research Methodology 

Tabgroup

Abstract

News recommendation systems aim to provide readers with new content based on previous reading behaviour and simultaneously allow them to discover new points of view by presenting them similar articles of different news sources. An important aspect of such content-based recommendation systems is the ability to recognize key news events within an article and link those events to other articles. The ENCORE project aims to lay the groundwork for a Dutch content-based recommendation algorithm by employing advances in Natural language processing (NLP) to perform event coreference resolution. The goal of event coreference resolution is to determine which mentions of events in texts refer to the same real-world event. The primary goal of this project is the coreference resolution between unrestricted events i.e events not belonging to a predetermined set of themes or topics in cross-document settings.

People

Supervisor(s)

Co-supervisor(s)

Phd Student(s)