Combining context and existing knowledge when recognizing biological entities - Early results

Abstract

Entity recognition has been studied for several years with good results. However, as the focus of information extraction (IE) and entity recognition (ER) has been set on biology and bioinformatics, the existing methods do not produce as good results as before. This is mainly due to the complex naming conventions of biological entities. In our information extraction system for biomedical documents called OAT (Ontology Aided Text mining system) we developed our own method for recognizing the biological entities. The difference to the existing methods, which use lexicons, rules and statistics, is that we combine the context of the entity with the existing knowledge about the relationships of the entities. This has produced encouraging preliminary results. This paper describes the approach we are using in our information extraction system for entity recognition. © 2008 Springer-Verlag Berlin Heidelberg.