N.G. Yarushkina, A.M. Namestnikov
Any CAD-document in intellectual repository should be saved in navigation structures for the organization of the publication of the document and the full-text query search. For the solution of this task it is necessary to construct an index for each document, to execute a set of documents clustering, to realize identification of problem area for the CAD-document. Indexing of the CAD-documents can be executed at level of terms or at level of concepts which are defined in domain ontology. Real storages of documents differ in considerable volume, a variety in terminology use, an unsharpness of borders of each class of subject domain. Therefore use of methods of soft computing for the solution of a problem of a CAD-documents clustering is represented reasonable. For the solution of this problem a neural network and genetic algorithm is adapted. The program system of intellectual storage of CAD-documents consists of the following subsystems: a subsystem of indexing, a clustering subsystem on the basis of a neural network, a clustering subsystem on the basis of Fuzzy-C-Means method, the qualifier and a clustering subsystem on the basis of genetic algorithm. Properties of fuzzy compliances are put in a basis of formal criterion of quality of domain ontology: fuzzy functionality, fuzzy injection fuzzy everywhere definiteness and fuzzy surjection.