Study On Model Resource Semantic Representation And Application

In organizational decision support system, decision-making formulation cannot leave the decision models. Data and models are widely distributed in the entire organization. It is a premise of implementing models sharing that the model resource should be found promptly and accurately. Model resource's semantic representation and semantic-based publishing and querying are two inseparable aspects in the model management process.In order to form unified understanding when users with different application background discuss the same model, semantic that the user required and the model resource implied should be represented explicitly and be understood by the computer. Ontology, as an important description tool of knowledge and concept, can express concepts' intention in an explicit and consistent way.The author attempted to employ ontology to provide support to semantic representation and processing in model management. Based on analysis of related research, algorithm of concept subsumption in ontology and question of model resource semantic annotation were discussed. Ontology-based model publishing and querying system was investigated. Model resource semantic representation and application was studied. Preliminary analysis and research was done on model resources semantic representation and application.The main works of this thesis are as follows:(1) Through analysis of the models that needed by the organizational decision support system, model resource mathematical model was established. Architecture of model resource semantic representation and application system was proposed.(2) Extension was made in the work of Tableau concept subsumption algorithm: Subsumptions of all concepts defined in the entire Tbox were calculated. LCSs (Least Common Subsumer) of concepts were calculated. Those concepts that subsume some concept but are not LCSs, especially those concepts that had the relation of grandparent and grandchild are considered, too.(3) The principle that annotation and the model resource itself should be separated is proposed, which overcomes the shortcoming that model resource itself had to be modified and makes the process of model resource annotation more flexible.(4) The concept of pre-calculation was proposed: during model publishing phase, subsumption relation between every concept in related ontology and published models is calculated. Publishing model is a one-time event but querying model is happened more often. Some time was spent in dealing with model semantic during the publishing phase. When model is queried, search is merely needed but calculation is unnecessary in getting queried model, hence response speed is improved.