Semantic Similarity Measures And Its Use In Design Management System

Semantic similarity has for a long time been a subject of intense scholarship in the fields of Artificial Intelligence and Psychology. The computational models in the field of Artificial Intelligence build upon a set of assumptions about similarity, which are tied to particular knowledge model. They are try to compute similarity from specific Knowledge Representation system. Meantime, the models in the field of Psychology are underlied by a wealth of experimental data, and try to imitate the aspects of human perception of similarity. Artificial intelligence researchers will be benefited from psychological achievements. On the basis of psychological studies about similarity, we propose a model, called the fuzzy contrast model, to measure the semantic similarity between concepts expressed by OWL DL.There are three key factors in measuring concept's similarity, concept, semantics of the concept and the similarity model. After thoroughly studying OWL DL and some related concepts of Description Logics, we transform an OWL DL concept to a set of axioms in Description Logic SHOIN(D) by the equivalence between OWL DL and SHOIN(D). The set of axiom is then transformed to a Structure Subsump-tion Normal Form(SSNF), and Explicit-Inclusion Items(ECItem), Implicit-Inclusion Items(ICItem), Role-Restricted Concepts(RCConcept) are extracted from the SSNF. It was proposed that the feature set of an OWL DL concept is made up of the ECItems, RCConcepts and ICItems of the concept. Their influence on the similarity measure was studied.FuzzyCon model is an extension to Rodriguez_Egenhofer model, which is an application of Tversky model. Fuzzy set was used to build the intersection set and difference of feature set in FuzzyCon model. The membership function of the fuzzy set is come from the similarity between concept in ontology. With the fuzzy set, FuzzyCon can represent individual view of sameness and difference for features. Feature weight was used to build salience function f in FuzzyCon model, and make the value of function f proportion to the quantity and weight of features in feature set. By the use of fuzzy set and feature weight, FuzzyCon model can perform personalized similarity...