The paper discusses improvement of approach to voice recognition for editing source code, using Google Cloud Speech. The improved approach combines neural networks with sound editing, editing distances, and replacement tables. The architecture for the recognition of Python language expressions is suggested. The results of the analysis of the testing approach on prototype program, that combines editing code on GitHub with Telegram, is discussed, The paper discusses advantages and disadvantages of the improved approach.

Keywords: voice recognition, neural networks, machine learning, source code analysis, formal languages, editing distances

The paper discusses approach to voice recognition, that allows to use tools of Google Cloud Speech platform for editing source code for programs, using audioprocessing, editing distances and substitution tables. The paper discusses issues of editing source code on mobile devices and the issues, which does not allow using tools from Google Cloud Speech platform directly, like not recognizing certain keywords. The paper suggests new method, which combines substitution tables and editing distances for solving the issues. This approach can be used for editing source code for programs, using mobile devices. The paper offers a prototypr of web-application, which allows to edit source code and uses this approach and, also, allows to submit changes to Github source code hosting platform and a popular instant messenger Telegram.

Keywords: voice recognition, mobile devices, source code analysis, formal languages, editing distances