Information system for converting audio Ukrainian-language text into written one based on NLP methods and machine learning

Abstract

Speech recognition provides various ways to analyse and process a user's recorded voice. It allows people to control different systems that are using one of the types of speech recognition. Speech-totext conversion is also a type of speech recognition that uses unscripted conversational data for further processing. This system involves several steps to process an audio file using electro-acoustic tools, sound filtering algorithms to find only relevant sounds, electronic datasets for the chosen language, and mathematical models that find appropriate words for a list of phonemes. People whose professions are related to typing large amounts of text using the keyboard can significantly speed up, facilitate the work process and reduce stress using systems that convert Speech-To-Text. In addition, such systems help businesses because remote work is becoming increasingly popular, and companies need tools for translating recorded audio from meetings to text for further analysis and systematization. The work's study object is converting audio in the Ukrainian language into its textual representation using NLP methods and machine learning. The scope of the research is audio file processing algorithms for finding relevant sounds and recognizing phonemes, as well as mathematical models for identifying words using an array of found phonemes. The work aims to design and develop an information system for converting audio in the Ukrainian language into its textual representation. According to the developments and calculations presented in work, namely: analysis of algorithms, areas of application and review of analogue problems in the first stage, the system analysis of the information system in the second stage and analysis and selection of relevant technologies and software development tools in the third stage, the information system for converting audio in the Ukrainian language into its textual representation was implemented in the form of a web application called «Ukrainian Speech-to-text», which is a technology for accurate and easy analysis of Ukrainian-language audio files and their other transcription into text. The application supports uploading files from the file system, recording them using a microphone, and saving the analysed data. The system is ready for use.