Abstract:A granular computing model for non-taxonomic relation learning from a domain information system was proposed.Firstly,an information function was extended and a domain information system was described formally;Secondly,one method of hierarchy granulating with respect to the domain information system was proposed.The method could generate information granules of different Abstract degrees;According to different domain concept granular spaces and the context analysis between concept granules,an approach to acquiring cross relationship between concept granules was proposed.The proposed model could provide one feasible method and technology of ontology non-taxonomic relations learning.