We describe the construction of a semantic database called SemCat consisting of a large number of semantically categorized names relevant to genomics. SemCat can be used to facilitate natural language processing in MEDLINE. We present suitable application areas including biomedical name classification and named entity recognition.