Event coreference detection in Estonian news articles: preliminary experiments

Abstract

Event coreference detection is a task of automatically determining which fine-grained textual descriptions of events (e.g. sentences) corefer. The task is important in organizing information in large collections of news articles, as extracted coreferring event mentions can provide the user with an overview of the events discussed in articles, and can also provide a glimpse into the factual data related to the events. In this article, we survey previous approaches to automatic event analysis, discuss theoretical considerations related to event coreference detection and also outline a motivation for experimenting with event coreference detection in the context of limited linguistic resources. In the experimental part of our work, we consider the task of event coreference detection in the subset of articles mentioning a specific person in a fixed publishing period, and we use the experiments to outline possible general factors that are influencing the results.

Keywords

• natural language processing
• text analysis
• Estonian