The Research Of The Matching And Integration Technology Of Biomedical Ontology

As a modeling tool which born from philosophy could describe concepts on both of semantic and knowledge layers, ontology could solve two important knowledge engineering problems: knowledge sharing and knowledge reusing. Recent years, as rapidly development of Life Science, a lot of ontologies are merging in the biomedical domain and many of the ontologies contain overlapping information. But these ontologies are often developed by different organizations and researchers and they are often heterogeneous in ontology representation and ontology buildings. To fully make advantage of the existing ontologies resources and integrating the resources efficiently, build special requirements oriented ontology base, we propose an ontology integration method based on task.The essential problem of ontology integration is the similarity measure of concept and relationship. Thereinto, concept similarity measure contains two aspects: (1) linguistic similarity (2) semantic similarity. Relationship similarity is measured by the concept similarity measure it related. The concepts of biomedical ontology are very complex and often comprised of phrase even phrase containing numbers. So, it is difficult to measure the similarity among biomedical concepts. To get better similarity results, firstly we preprocess the concept and break it into tokens and give every token a part of speech. Secondly, we combine WordNet and Edit distance similarity methods to measure the similarity between every pair of tokens from two concepts. Finally, we average the similarities which derived from the pair of tokens with weight and get better results in the linguist level. On the other hand, we take advantage of the structure of ontology and use the domain ontology UMLS to help for measuring the structure similarity of concepts and we get better results in the Semantic level. To take advantage of existing ontologies and build special requirement oriented ontology base, in this paper, we introduce core ontology and finish the integrating overlapping information by integrating core ontologies. The essence of core ontology is the sub-ontology of the original ontology. Core ontology is the concepts set which are nearly to the central concept and the relationships set which is related to the central concept. In this sense, core ontology is the view of task in the ontology. At last, we certified the method by experiments and they show that the ontology integration method based on task improve the efficiency and correction of integration.Finally, we explored a system of ontology integration based on task, and described the core modules and program's flow in the system.