A 2D Convolutional Gating Mechanism for Mandarin Streaming Speech Recognition

Wang, Xintong; Zhao, Chuangang

doi:10.3390/info12040165

Cited by 2 publications

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Organizing and conducting manual testing of Putonghua proficiency consumes a lot of human and material resources, and it is impossible to avoid the influence of subjective factors in scoring. Therefore, many researchers have carried out studies on objective testing of Putonghua proficiency with the help of computers, and a lot of progress has been made [1][2][3].…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Based Algorithm for Detecting Errors in Mandarin Read-Aloud Backreading Omission Incremental Reading

Li,

Yang

2024

Applied Mathematics and Nonlinear Sciences

View full text Add to dashboard Cite

With the continuous development of computer technology and the deepening of the popularization of Mandarin, the role of computer technology in language-assisted language learning and recognition has become more and more significant. In this study, the acoustic model and speech model based on the Hidden Markov Model are constructed for the error detection of reading aloud in Mandarin. Then the feature extraction of the acoustics of speech signals is carried out to build the error detection model of reading aloud in Mandarin based on pronunciation features. On this basis, a DNNHMM hybrid model is built by combining deep neural networks and Hidden Markov Models for detecting Mandarin read aloud keyword errors. Through the empirical analysis of Mandarin reading error detection, it can be seen that the average accuracy of the Mandarin reading error detection model constructed in this paper is 92.37%. Compared with the other models, the average accuracy improvement ranges from 4.69% to 8.19%, and the average accuracy of the vowel and consonant pronunciation features is 85.04% and 81.69%, respectively. In the detection of errors in Mandarin reading aloud, the F-score of misreading, back-reading, adding, changing, omitting, and other six error types is above 80%, and the accuracy rate is above 83%. It shows that the deep learning-based model for reading aloud in Mandarin performs well and provides an effective method for error detection in reading aloud in Mandarin.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep Learning Based Algorithm for Detecting Errors in Mandarin Read-Aloud Backreading Omission Incremental Reading

Li,

Yang

2024

Applied Mathematics and Nonlinear Sciences

View full text Add to dashboard Cite

show abstract

A Bidirectional Context Embedding Transformer for Automatic Speech Recognition

et al. 2022

View full text Add to dashboard Cite

Transformers have become popular in building end-to-end automatic speech recognition (ASR) systems. However, transformer ASR systems are usually trained to give output sequences in the left-to-right order, disregarding the right-to-left context. Currently, the existing transformer-based ASR systems that employ two decoders for bidirectional decoding are complex in terms of computation and optimization. The existing ASR transformer with a single decoder for bidirectional decoding requires extra methods (such as a self-mask) to resolve the problem of information leakage in the attention mechanism This paper explores different options for the development of a speech transformer that utilizes a single decoder equipped with bidirectional context embedding (BCE) for bidirectional decoding. The decoding direction, which is set up at the input level, enables the model to attend to different directional contexts without extra decoders and also alleviates any information leakage. The effectiveness of this method was verified with a bidirectional beam search method that generates bidirectional output sequences and determines the best hypothesis according to the output score. We achieved a word error rate (WER) of 7.65%/18.97% on the clean/other LibriSpeech test set, outperforming the left-to-right decoding style in our work by 3.17%/3.47%. The results are also close to, or better than, other state-of-the-art end-to-end models.

show abstract

A 2D Convolutional Gating Mechanism for Mandarin Streaming Speech Recognition

Cited by 2 publications

References 20 publications

Deep Learning Based Algorithm for Detecting Errors in Mandarin Read-Aloud Backreading Omission Incremental Reading

Deep Learning Based Algorithm for Detecting Errors in Mandarin Read-Aloud Backreading Omission Incremental Reading

A Bidirectional Context Embedding Transformer for Automatic Speech Recognition

Contact Info

Product

Resources

About