Voice Recognition Methods and Modules for the Development of an Intelligent Virtual Consultant Integrated with WEB-ERP

Abstract

There were explored the methods and modules of voice recognition for the development of an intelligent virtual consultant, which plays an important role as an auxiliary tool when working with business process information systems and more precisely with the ERP system in this article. This study is intended to characterize an intelligent virtual consultant in evaluating the usefulness and effectiveness of using specific functions in a simulated enterprise resource planning (ERP) business process software system. The intelligent virtual consultant is integrated with the Web-ERP prototype. The Mel-frequency-Cepstral coefficient algorithm was utilized in conjunction with the Levenberg-Marquardt (LM) and Broyden-Fletcher-Goldfarb-Shanno (BFGS) methods as the underlying basis for the analysis and extraction of features. A comparative analysis of these methods was also given to identify the effectiveness of used methods. A great benefit of this virtual consultant is the voice input of long text data fields when working on Web-ERP. This system is relevant because it significantly expands the search for business workflows. The developed system makes it possible to convert speech into text by extracting information system instructions. These texts are passed to an ontological database in which a query is executed using terms in order to get a set of available commands for its execution. When implementing an intelligent virtual consultant, it increases the capabilities of the application using not only user navigation through the system, but also makes it possible to navigate and explain data using speech synthesis. This prototype developed in the research will be used in a combination of an intelligent virtual consultant and data analysis.