A study of transformer-based end-to-end speech recognition system for Kazakh language

Abstract

Today, the Transformer model, which allows parallelization and also has its own internal attention, has been widely used in the field of speech recognition. The great advantage of this architecture is the fast learning speed, and the lack of sequential operation, as with recurrent neural networks. In this work, Transformer models and an end-to-end model based on connectionist temporal classification were considered to build a system for automatic recognition of Kazakh speech. It is known that Kazakh is part of a number of agglutinative languages and has limited data for implementing speech recognition systems. Some studies have shown that the Transformer model improves system performance for low-resource languages. Based on our experiments, it was revealed that the joint use of Transformer and connectionist temporal classification models contributed to improving the performance of the Kazakh speech recognition system and with an integrated language model it showed the best character error rate 3.7% on a clean dataset.

Rapid quantitative authentication and analysis of camellia oil adulterated with edible oils by electronic nose and FTIR spectroscopy

“…The model is built on an encoder-decoder architecture, each consisting of multiple layers. Notably, the Transformer eliminates the need for recurrence, which is a staple in traditional models such as recurrent neural networks, thereby facilitating enhanced parallelization ( Orken et al, 2022 ). The architecture incorporates a multi-head self-attention mechanism and a position-wise fully connected feed-forward network in both the encoder and decoder layers.…”

Automatic Speech Recognition for Uyghur, Kazakh, and Kyrgyz: An Overview

“…In 2016, Abilhayer et al [141] constructed a continuous speech recognition system based on the GMM-HMM. Since 2019, the Institute of Information and Computational Technology and al-Farabi Kazakh National University published several papers on Kazakh speech recognition and verified the DNN-HMM system [142,143], BLSTM-CTC end-to-end system [144,145], and Transformer CTC/attention system [146] with their private data. At the same time, Beibut et al [147] constructed an LSTM-CTC end-to-end Kazakh ASR system based on transfer learning.…”

Implementation of artificial intelligence in speech (voice input) in the conditions of a high-flow endoscopy

“…документации; • голосовое управление (использование команд); • интерактивное взаимодействие с пациентом. В Казахстане были произведены испытания системы распознавания речи Trasformer, которая показала хорошие результаты (коэффициент ошибок 3.7%) [3]. Автоматическое распознавание речи широко используется на английском, китайском, японском и французском языках.…”