Double Layered Learning for Biological Event Extraction from Text.

This paper presents our approach (referred to as BioEvent) for protein-level complex event extraction, developed for the GENIA task (Kim et al., 2011b) of the BioNLP Shared Task 2011 (Kim et al., 2011a). We developed a double layered machine learning approach which utilizes a state-of-the-art minimized feature set for each of the event types. We improved the best performing system of BioNLP 2009 overall, and ranked first amongst 15 teams in finding "Localization" events in 2011. BioEvent is available at http://bioevent.sourceforge.net/