Grammatical inference in practice: A case study in the biomedical domain

In this paper we discuss an approach to named entity recognition (NER) based on grammatical inference (GI). Previous GI approaches have aimed at constructing a grammar underlying a given text source. It has been noted that the rules produced by GI can also be interpreted semantically [16] where a non-terminal describes interchangeable elements which are the instances of the same concepts. Such an observation leads to the hypothesis that GI might be useful for finding concept instances in a text. Furthermore, it should also be possible to discover relations between concepts, or more precisely, the way such relations are expressed linguistically. Throughout the paper, we propose a general framework for using GI for named entity recognition by discussing several possible approaches. In addition, we demonstrate that these methods successfully work on biomedical data using an existing GI tool.