EXPERT SYSTEMS USING FOR ANSWERS ANALYSIS IN AUTOMATED KNOWLEDGE CONTROL SYSTEMS

Abstract

This article proposes the use of an expert system to analyze answers to test questions of different types: open questions, closed questions, questions for compliance, questions for the correct sequence. The Cartesian distance between answers, Levenshtein distances and the relative number of errors were used to build the computer mathematical model. The models of an expert system and logical inference micromachines are considered. The proposed model does not carry out a direct dialogue with the testee (user). Interaction with the testee (user) will be carried out through the automated knowledge control system interface and its database - analysis of answers, generation of additional questions, saving the progress of the inference (solution), etc. In order to speed up development and facilitate the expansion of the system's functionality, we  implemented the logical inference machine in the form of a controller and several logical inference micromachines. The inference machine analyzes the testee answers and if necessary, it generates additional questions for him or marks his answers as guessed. Each of the logical inference micromachines solves only one task and can be launched a limited number of times within the same testing session. The described expert system makes the testing process more similar to the procedure of interaction between a teacher and a student, allowing to clarify or discard the received answers. The proposed expert system and logical inference micromachines were implemented as the software module for automated knowledge control system “Antonov Students Test System (A.S.T.S)” and was used in practice more than 5 years. As a result of the operation of the logic inference machine 1, it was found that 4.59% of the answers received needed clarification. From all of the re-asked questions 20.68% were able to get completely correct answers and thus improve the result. As a result of the operation of the logic inference micromachine 2, the following patterns were revealed: 60.54% are for pairs who do not need to clarify the results; 25.14% are pairs of answers, according to which the expert system made a decision to guess the answer (5.24% of the total number of answers). As a result of the operation of inference micro-machines 1 and 2, 6.19% of the total number of answers were changed.