-
Humanities and the arts
- Computational linguistics
- Pragmatics
We humans are flooded with new facts and events every day We communicate about this through references, by using synonyms and other such strategies These are semantic techniques that facilitate smooth discourse Easy communication is something that computers still struggle with today In this project, we hope to change this and help computers understand texts We do this by focusing on the issue of coreference resolution (linking different mentions of the same entities or events to each other) We want to design a system that can establish coreference links within one text, but also across text boundaries
We will work on news events in Flemish news articles In doing so we do not restrict ourselves to using a limited typology of events Our goal is to find all mentions of events (event detection) and to link them (event coreference) This is a task that has hardly been investigated for Dutch and which is useful for many applications Within this project, we will investigate how news event coreference can contribute to better, more diverse news recommendation