UDC 004.912
INTEGRATION OF MULTI-AGENT AND NEURAL NETWORK TECHNOLOGIES IN A PATTERN RECOGNITION SOFTWARE SYSTEM FOR ROBOTIC DOCUMENT FLOW
A. V. Krivosheev, graduate student of Samara State Technical University, Samara, Russia;
orcid.org/0000-0002-9664-1317, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
The problem of integrated application of software development technologies with the elements of artificial intelligence to expand the possibilities of their use as part of modern robotic document flow systems is considered. A method of multi-agent ensemble of intelligent components of an adaptive pattern recognition system is proposed, which consists of encapsulating artificial neural networks through software agents, united through links and connections into a virtual world architecture that most rationally ensures their interaction with each other to implement automatic dispatch and competition strategies. Testing of the method and architecture proposed was also carried out in the implementation of intelligent system for text comprehension and text generation for the joint use of several artificial neural networks with different training data sets for a comprehensive solution to various problems of semantic analysis of texts in Russian. The use of ensemble of intelligent components within the framework of dynamic dispatch makes it possible to increase the performance of pattern recognition software as part of a document management system by 2.5 times compared to classical methods of combining neural networks
Key words: : pattern recognition, text processing, software system architecture, artificial intelligence technologies, machine learning, multi-agent technologies, ensemble.