Domain ontology and top-level ontology: how can we co-ordinate the two?

Summary form only given. The author discusses a specific domain ontology, GENIA ontology, which consists of 32 semantic categories in life science field. The domain ontology is established to annotate papers in the field semantically, and has been use to annotate 2000 abstracts (around 500,000 words). The ontology is now being extended to cover other semantic categories like diseases, symptoms of diseases, body parts, etc. the categories that are not the major ones in the genome science field but, nonetheless frequently appear in papers of the field. In order to extend the GENIA ontology to cover these categories, we have exploited existing domain ontologies like UMLS and the GENE ontology. Our effort is a typical example of ontology mapping among domain ontologies constructed by different groups of domain specialists. The next crucial extension of the GENIA ontology is to cover concepts related with events, which involves mappings of verb subcategorization frames with distinct events of the domain. We assume that the mapping will be n to n and that the event ontology will require much finer semantic distinctions than those of verb subcategorization frames or distinctions of a general ontology. We are particularly interested in how we can use the general ontology of events to accelerate acquisition of event ontology specific to the domain.