Ajou University repository

Anti-focal loss for speech recognition on small-scale datasets
Citations

SCOPUS

0

Citation Export

Publication Year
2021-08-20
Journal
2021 4th International Conference on Pattern Recognition and Artificial Intelligence, PRAI 2021
Publisher
Institute of Electrical and Electronics Engineers Inc.
Citation
2021 4th International Conference on Pattern Recognition and Artificial Intelligence, PRAI 2021, pp.19-22
Keyword
Few-shot learningSmall-scale dataSpeech recognitionTransformer
Mesh Keyword
Automatic speech recognition systemEncoder-decoder architectureFew-shot learningLearning modelsPrediction tasksSequential predictionSmall scaleSmall-scale dataTransformerTransformer modeling
All Science Classification Codes (ASJC)
Artificial IntelligenceComputer Vision and Pattern Recognition
Abstract
Deep learning models with encoder-decoder architecture become popular in automatic speech recognition systems, due to their success in sequential prediction tasks. Recently, the conformer model has greatly improved the accuracy of speech recognition. However, similar to transformer models, its training relies on a large amount of data. This paper explores an efficient few-shot learning strategy. Specifically, a spec-augment approach is proposed to augment the speech dataset, then a novel loss function, anti-focal loss, is introduced to encourage fast convergence in a small-scale, unbalanced data setting. Extensive experiments on aishell-l dataset show that our model outperforms state-of-the-art approaches under limited support data, in terms of convergence speed and generalization ability.
Language
eng
URI
https://aurora.ajou.ac.kr/handle/2018.oak/36714
https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85117960182&origin=inward
DOI
https://doi.org/10.1109/prai53619.2021.9550804
Journal URL
http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=9550757
Type
Conference
Show full item record

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Related Researcher

Chung, Tae-Sun Image
Chung, Tae-Sun정태선
Department of Software and Computer Engineering
Read More

Total Views & Downloads

File Download

  • There are no files associated with this item.