1. PROCEDURE FOR DEBUGGING REPLICA CLASS ERRORS IN RULE-ORIENTED KNOWLEDGE BASES
- Author
-
Serhii Osiievskyi, Oleksii Kolomiitsev, Viacheslav Tretiak, Oleksii Petrenko, Olha Petrenko, Andrii Vlasov, Ruslan Oliinyk, Yurij Zhivetc, Oleksandr Shumigay, and Pavlo Baturin
- Abstract
The problems of error detection by experts in the rule-oriented knowledge bases of oriented information systems, which occurring at the stage of the knowledge base formation, are considered. It has been determined that such errors are associated with the contradictory opinions of experts and / or the limited (imperfect) description of the subject area. The approaches to their debugging were analyzed. The ways of improving the existing approaches to debugging static errors of the “replica” class and possible ways of applying the obtained solutions to eliminate errors “contradictoriness”, “redundancy”, “incompleteness” have been shown. Considering the requirements for the accuracy of information presentation, the solutions for expanding the content of the definition of the static error of the rule-oriented knowledge base have been proposed. The issues of the influence of the "replica" class errors on the results of output in accordance with the rules of the rule-oriented knowledge base were considered. The possibility of using graph theory methods for solving the problem of debugging errors of the "replica" class has been proved. The algorithmic structures for identifying and debugging errors of the specified class, which in contrast to existing solutions allow to identify duplicate apexes at each rank of the graph to which the rule-oriented knowledge base is amounted to, have been developed. A software implementation of detecting and debugging static errors of incomplete, fragmentary and complete duplication has been developed. Due to usage of recursion, the requirements for preparing an array of input data for processing were reduced. The solutions obtained are compatible with the requirements of DSTU ISO / IEC 9126, DSTU ISO / IEC 14598 and consider the requirements of the Software Quality Requirements and Evaluation family of standards as the value of the apexes of the event tree graph. In the process of solving the problem, the specifics of the functioning of the man-machine system was considered. In particular, the possibility of formalizing various aspects of knowledge (alethic, dissisional, casual, dianistic) and assurance of a given level of operativeness in finding solutions has been accounted.
- Published
- 2022
- Full Text
- View/download PDF