An empirical study of a query-language front-end processor using BK-products for fuzzy relational knowledge-based systems.

摘要

Zadeh, the father of fuzzy logic, recently initiated a new field--computing with words (CW)--which provides a conceptual framework for computing with words rather than numbers. Our work addresses the advantages associated with fuzzy relations used in building systems for computing with words. Bandler and Kohout introduced new non-associative products of fuzzy relations (called BK-products in the literature). Such products are used to build architectures for fuzzy relational computing in AI and other applications. In BK-relational architectures, the problems and solutions are expressed in form of fuzzy mathematical formulas.; A prototype front-end unit is developed for translating linguistic queries into BK-product formulas as well as for back-translation of BK-product formulas into linguistic statements. Thus our work makes it possible to convert BK-architectures to special kind of systems for computing with words. An English Query Language (EQL) grammar is introduced to allow a suitably large variety of queries to be accepted by the front-end prototype. The translation of EQL query requests as well as the back-translation is processed incrementally involving syntax and semantic phases. Our mechanism for translation from EQL to relational formulas as well as the back-translation take the advantage of the fact all relational formulas can be combined with semiotic descriptors in BK relational architectures. The fuzzy relations found in the prototype's sample data dictionary were created with sample information obtained from CLINAID--a medical knowledge based system using BK-products of relations for fuzzy interval inference. Our queries are not just unqualified sentences but linguistic qualifiers are also employed in EQL. An important feature of this system prototype is that it can be used as an empty shell for creating "computing with words" applications for other knowledge domains such as industrial engineering and computer science e.g. in fuzzy information retrieval, or systems for verification of trustworthiness and security of computer systems.