Book chapter · CRC Press eBooks · June 27, 2024

Language and acoustic modeling in Uzbek speech recognition

Nilufar A’loxanovna Niyozmatova, Narzillo Mamatov, Abdurashid Samijonov, B.N. Samijonov

Abstract

This paper is devoted to the evaluation of speech recognition (NTO) approaches, in which the general structure of automatic speech recognition algorithms, language and acoustic modeling, and lexical data are analyzed. In the conducted experiments, several existing applications of speech recognition contrasted using Uzbek-language speech test selections as the basis. The choices encompass four distinct NATO systems employing diverse acoustic and language modeling algorithms. Only two of these systems employed a consistent approach across all components, while in the other two instances, the acoustic and language models were founded on disparate algorithms, resulting in fundamentally different structures for all elements within the selection. Data sets are analyzed by each system using Python programming language applications. The test results were analyzed and conclusions were drawn that the effectiveness of the automatic speech recognition system depends on the optimization of its elements and training using the required data set, neural network and statistical approaches are equally effective in language and acoustic modeling.

Speech Recognition and SynthesisSpeech and Audio ProcessingSpeech and dialogue systemsUzbek languagespeech recognitionlanguage modeling

Language and acoustic modeling in Uzbek speech recognition

Abstract

More by Narzillo Mamatov