Open AccessThe article considers a corporate knowledge base representing a set of the large number of various semi-structured documents, which describe to some detail precedents i.e. some situations and decisions made in these situations. The task is to find the decision in such knowledge base and search the most suitable precedents and their corresponding documents in it.
The work continues a series of the authors’ works, which develop an original approach to the solution of this task. It is supposed that metadata of the knowledge base documents are based on the ontology of the corresponding subject domain specified as a semantic network. As opposed to the previous works it is necessary that ontology of subject domain, and thereby and corresponding semantic network, is hierarchically clustered based on the roles of concepts.
Documents in the knowledge base, and also search queries are presented as frames, which are called ‘patterns of design and inquiry’, respectively. Slots of these patterns correspond to roles of concepts of the used ontology. Above-noted roles break concepts of ontology, document, and request to the knowledge base into clusters. It is supposed that semantic networks of the abovementioned clusters are designed by a technique for creation of a semantic network of the document. Thus, search images of the document and inquiry are presented as a set of the semantic networks corresponding to slots of a pattern of design and a pattern of inquiry.
The work offers the model of a semantic network of the ontological knowledge base using a hierarchical role clustering of concepts of this ontology. The task for a multi-criteria evaluation of relevance of corporate ontological knowledge base documents using their hierarchical role clustering is set. The method is offered to solve the task using the original measures of proximity of role patterns of design of the required document and inquiry.