Biomedical event detection using rules, conditional random fields and parse tree distances

Sarafraz, F., Eales, J., Mohammadi, R., Dickerson, J., Robertson, D. and Nenadic, G. (2009) Biomedical event detection using rules, conditional random fields and parse tree distances. In: Tsujii, J.'i. (ed.) Proceedings of the Workshop on BioNLP: Shared Task. Series: ACM conference proceedings series. Association for Computational Linguistics: Morristown, N.J., pp. 115-118.

Abstract

This paper reports on a system developed for the BioNLP'09 shared task on detection and characterisation of biomedical events. Event triggers and types were recognised using a conditional random field classifier and a set of rules, while event participants were identified using a rule-based system that relied on relative distances between candidate entities and the trigger in the associated parse tree. The results on previously unseen test data were encouraging: for non-regulatory events, the F-score was almost 50% (with precision above 60%), with the overall F-score of around 30% (49% precision). The performance on more complex regulatory events was poor (F-measure of 7%). Among the 24 teams submitting the test results, our results were ranked 12th for the overall F-score and 8th for the F-score of non-regulation events.

• Item Type: Book Sections
• Status: Published
• Refereed: Yes
• Glasgow Author(s) Enlighten ID: Dickerson, Dr Jonathan
• Authors: Sarafraz, F., Eales, J., Mohammadi, R., Dickerson, J., Robertson, D., and Nenadic, G.
• College/School: College of Medical Veterinary and Life Sciences > School of Medicine, Dentistry & Nursing
• Journal Name: Proceedings of the Workshop on Current Trends in Biomedical Natural Language Processing
• Publisher: Association for Computational Linguistics