Computer linguistic approach for computational creativity tasks

Abstract

At present, information automated systems for finding technical solutions contain a module that allows calculating the values of requirements for synthesized models of the principle of operation to the design object, in an automatic mode. Traditionally, the values of these requirements are presented in the form of quantitative estimates. To extend the capabilities of information automated systems, it is proposed to use requirements whose values are represented as a linguistic variable that is considered given if all elements of the tuple are defined, including the name of the linguistic variable. This problem is complicated by the fact that for different types of representation of input quantities (qualitative - quantitative) as a result of the application of arithmetic and logical operations, the functions of the resulting fuzzy sets can be obtained, which are difficult to identify. The paper proposes a technique that allows the values of the membership function to be associated with their semantic meaning for the representation of the linguistic variable. To solve this problem, the neural network with adaptive resonance feature is used: to retain plasticity when memorizing new images, and at the same time to prevent the modification of old memory. It is shown that parallel computations of a neural network occur faster and do not increase with the increase in the number of effects models in the chain of the model of the operation principle of the design object. The developed information automated search system for technical solutions covering the tasks of calculating requirements, both with clear and linguistic values, is applied at the stage of search design of technical objects. The tasks that are considered at this stage relate to the tasks of TSIP (the Theory of solving inventive problems). The use of an automated system for searching technical solutions is aimed at supporting scientists and inventors to quickly find solutions to creative problems in the design of technical objects.