Loading [a11y]/accessibility-menu.js
An Improved Speech Recognition System Based on Transformer Language Model | IEEE Conference Publication | IEEE Xplore

An Improved Speech Recognition System Based on Transformer Language Model


Abstract:

With the application of deep learning in speech recognition technology, many neural network-based model structures have emerged in speech recognition, while as another im...Show More

Abstract:

With the application of deep learning in speech recognition technology, many neural network-based model structures have emerged in speech recognition, while as another important part of language models, traditional statistical language models have not been further developed due to drawbacks such as oversized parameters, purely based on statistical frequencies, and poor generalization ability. To address those drawbacks, this paper uses a fully convolutional deep neural network-based acoustic model, which is then combined with a connected temporal classification method to replace the traditional statistical language model with a Transformer-based language model. Experiments on several datasets such as THCHS30 show that the recognition effectiveness of the Transformer-based language model speech recognition system is effectively improved.
Date of Conference: 14-17 October 2021
Date Added to IEEE Xplore: 09 December 2021
ISBN Information:

ISSN Information:

Conference Location: Xi'an, China

References

References is not available for this document.