| Ontology is a concept model that can describe information in semantic level andknowledge level,is mainly used to describe the concepts and the relations between theconcepts.Since is been proposed,has attracted many researchers ’attention in domesticand foreign,and have been widely applied in computer and many other fields,such asknowledge management,information retrieval, semantic Web, electronic commerce andintelligent information systems integration and other aspects.Although the existingediting environment of ontology construction tools can satisfy the demand ofestablishing Ontology,but relying entirely on the artificial collection of relationshipsbetween concepts to construct ontology,is still a fee to work hard.In order to use theknowledge acquisition techniques to reduce the overhead of ontology construction,thenusing the ontology learning.Ontology learning combines ontology engineering, machinelearning and statistical method of automatic or semi-automatic to construct theontology.In recent years, ontology learning becomes a hot topics in current research.Ontology learning tasks including concept acquisition,relation acquisition, axiomacquisition.These three aspects constitute the ontology learning from difficult to easylevels.This paper mainly studies the acquisition methods of relationship of concepts inontology learning,including the taxonomic relations and the non-taxonomicrelations.The main research work is as follows:1) Presents the general framework of ontology prototype system, has done adetailed analysis on demand and the key technology of the prototype of system.2) Respectively based on pattern matching and clustering method to realize thetaxonomic relation acquisition of concepts, the implementation of clustering, presentsthe improved K-means clustering method to obtain a taxonomic relationship,and made comparison and analysis on two realizing methods.3) As to the non-taxonomic relations, first use of the expansion of the associationrules and heuristic-based of AE, then presents based on VF*ICF and Logarithmiclikelihood ratio method to get non-taxonomic relations, at last made analyzed andcompared of three methods.4) Finally we implement a prototype system for ontology learning to made theimplementation of relations acquisition combining with the above research methods. |
