KEYWORD-BIASED SPEECH RECOGNITION: A COMPARATIVE STUDY-Reference-Cited by-同舟云学术

KEYWORD-BIASED SPEECH RECOGNITION: A COMPARATIVE STUDY

Published:2023 Issue: Volume: Page:
ISSN:
Container-title:Proceedings
language:
Short-container-title:

Author:

KARAMYAN D.S.,KIRAKOSYAN G.A.

Abstract

The ability to recognize unknown or rare words, such as technical terms and names, is crucial for speech recognition technology to accurately comprehend conversations in context. Nonetheless, current end-to-end speech recognition models often have difficulty in recognizing words that are rarely or never seen during training. This paper examines the effectiveness of different keyword biasing methods, as well as their modifications outlined in previous studies, which do not require any modifications to the ASR model.

Publisher

National Polytechnic University of Armenia

Reference21 articles.

1. Chan W., Jaitly N., Le Q., and Vinyals O. Listen, attend and spell: A neural network for large vocabulary conversational speech recognition // IEEE international conference on acoustics, speech and signal processing (ICASSP). -2016. -P. 4960–4964.

2. Robust speech recognition via large-scale weak supervision / A. Radford, J.W. Kim, T. Xu, G. Brockman, et al: ArXiv preprint arXiv:2212.04356, 2022.

3. Graves A., Fernández S., Gomez F., and Schmidhuber J. Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks // Proceedings of the 23rd International Conference on Machine Learning. -2006. -P. 369–376.

4. Graves A. Sequence transduction with recurrent neural networks: ArXiv preprint arXiv:1211.3711, 2012.

5. Deep context: end-to-end contex¬tu¬a¬l speech recognition/ G. Pundak, T.N. Sainath, R. Prabhavalkar, et al // IEEE spoken language technology workshop (SLT). -2018. -P. 418–425.