Fast text-only domain adaptation of RNN-transducer prediction network

Show full item record



Permalink

http://hdl.handle.net/10138/336639

Citation

Pylkkonen , J , Ukkonen , A , Kilpikoski , J , Tamminen , S & Heikinheimo , H 2021 , Fast text-only domain adaptation of RNN-transducer prediction network . in 22nd Annual Conference of the International Speech Communication Association, INTERSPEECH 2021 . Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH , vol. 2 , ISCA , Baixas , pp. 1882-1886 , Annual Conference of the International Speech Communication Association , Brno , Czech Republic , 30/08/2021 . https://doi.org/10.21437/Interspeech.2021-1191

Title: Fast text-only domain adaptation of RNN-transducer prediction network
Author: Pylkkonen, Janne; Ukkonen, Antti; Kilpikoski, Juho; Tamminen, Samu; Heikinheimo, Hannes
Other contributor: University of Helsinki, Department of Computer Science
Publisher: ISCA
Date: 2021
Language: eng
Number of pages: 5
Belongs to series: 22nd Annual Conference of the International Speech Communication Association, INTERSPEECH 2021
Belongs to series: Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
ISBN: 9781713836902
DOI: https://doi.org/10.21437/Interspeech.2021-1191
URI: http://hdl.handle.net/10138/336639
Abstract: Adaption of end-to-end speech recognition systems to new tasks is known to be challenging. A number of solutions have been proposed which apply external language models with various fusion methods, possibly with a combination of two-pass decoding. Also TTS systems have been used to generate adaptation data for the end-to-end models. In this paper we show that RNN-transducer models can be effectively adapted to new domains using only small amounts of textual data. By taking advantage of model's inherent structure, where the prediction network is interpreted as a language model, we can apply fast adaptation to the model. Adapting the model avoids the need for complicated decoding time fusions and external language models. Using appropriate regularization, the prediction network can be adapted to new domains while still retaining good generalization capabilities. We show with multiple ASR evaluation tasks how this method can provide relative gains of 10-45% in target task WER. We also share insights how RNN-transducer prediction network performs as a language model.
Subject: Adaptation
Automatic speech recognition
End-to-end models
Language model
RNN-transducer
113 Computer and information sciences
6121 Languages
Rights:


Files in this item

Total number of downloads: Loading...

Files Size Format View
pylkkonen21_interspeech.pdf 228.6Kb PDF View/Open

This item appears in the following Collection(s)

Show full item record