Output details
11 - Computer Science and Informatics
University of Manchester
Boosting automatic event extraction from the literature using domain adaptation and coreference resolution
<24> Extracting complex relations (events) from text is important for advanced fine-grained semantic search. Challenges for event extraction systems are generalization to new domains and event types, and coreference resolution (linking event entities across text). The novelty here is incorporation of these abilities in EventMine, making it the best performing system in extracting events from biomedical text (evidence: comparative evaluation in community shared tasks (BioNLP)). EventMine underpins event extraction for Europe PubMed Central text-mining based search facilities (having extracted 80m events from 2.5m documents), and two search services offered by the National Centre for Text Mining.