The Relevance of the Source Language in Transfer Learning for ASR
This study presents new experiments on Zyrian Komi speech recognition. We use Deep-Speech to train ASR models from a language documentation corpus that contains both contemporary and archival recordings. Earlier studies have shown that transfer learning from English and using a domain matching Komi...
Published in: | Proceedings of the Workshop on Computational Methods for Endangered Languages |
---|---|
Main Authors: | , , , |
Other Authors: | , |
Format: | Conference Object |
Language: | English |
Published: |
2022
|
Subjects: | |
Online Access: | http://hdl.handle.net/10138/351332 |
id |
ftunivhelsihelda:oai:helda.helsinki.fi:10138/351332 |
---|---|
record_format |
openpolar |
spelling |
ftunivhelsihelda:oai:helda.helsinki.fi:10138/351332 2024-01-07T09:44:40+01:00 The Relevance of the Source Language in Transfer Learning for ASR Hjortnæs , Nils Partanen, Niko Rießler, Michael Tyers, Francis M. The National Library of Finland, Library Network Services Department of Finnish, Finno-Ugrian and Scandinavian Studies 2022-12-02T08:07:03Z 7 application/pdf http://hdl.handle.net/10138/351332 eng eng 10.33011/computel.v1i.959 Proceedings of the 4th Workshop on the Use of Computational Methods in the Study of Endangered Languages Proceedings of the Workshop on Computational Methods for Endangered Languages 978-1-954085-01-5 Hjortnæs , N , Partanen , N , Rießler , M & Tyers , F M 2021 , The Relevance of the Source Language in Transfer Learning for ASR . in Proceedings of the 4th Workshop on the Use of Computational Methods in the Study of Endangered Languages . vol. 1 , Proceedings of the Workshop on Computational Methods for Endangered Languages , The Association for Computational Linguistics , pp. 63-69 , Workshop on the Use of Computational Methods in the Study of Endangered Languages , 02/03/2021 . https://doi.org/10.33011/computel.v1i.959 conference ORCID: /0000-0001-8584-3880/work/118364555 ORCID: /0000-0002-2397-2860/work/118365886 f93ccf3d-0870-447c-a825-15a0b3524679 http://hdl.handle.net/10138/351332 unspecified openAccess info:eu-repo/semantics/openAccess 6121 Languages Conference contribution publishedVersion 2022 ftunivhelsihelda 2023-12-14T00:13:17Z This study presents new experiments on Zyrian Komi speech recognition. We use Deep-Speech to train ASR models from a language documentation corpus that contains both contemporary and archival recordings. Earlier studies have shown that transfer learning from English and using a domain matching Komi language model both improve the CER and WER. In this study we experiment with transfer learning from a more relevant source language, Russian, and including Russian text in the language model construction. The motivation for this is that Russian and Komi are contemporary contact languages, and Russian is regularly present in the corpus. We found that despite the close contact of Russian and Komi, the size of the English speech corpus yielded greater performance when used as the source language. Additionally, we can report that already an update in DeepSpeech version improved the CER by 3.9% against the earlier studies, which is an important step in the development of Komi ASR. Peer reviewed Conference Object Komi language HELDA – University of Helsinki Open Repository Proceedings of the Workshop on Computational Methods for Endangered Languages 1 2 |
institution |
Open Polar |
collection |
HELDA – University of Helsinki Open Repository |
op_collection_id |
ftunivhelsihelda |
language |
English |
topic |
6121 Languages |
spellingShingle |
6121 Languages Hjortnæs , Nils Partanen, Niko Rießler, Michael Tyers, Francis M. The Relevance of the Source Language in Transfer Learning for ASR |
topic_facet |
6121 Languages |
description |
This study presents new experiments on Zyrian Komi speech recognition. We use Deep-Speech to train ASR models from a language documentation corpus that contains both contemporary and archival recordings. Earlier studies have shown that transfer learning from English and using a domain matching Komi language model both improve the CER and WER. In this study we experiment with transfer learning from a more relevant source language, Russian, and including Russian text in the language model construction. The motivation for this is that Russian and Komi are contemporary contact languages, and Russian is regularly present in the corpus. We found that despite the close contact of Russian and Komi, the size of the English speech corpus yielded greater performance when used as the source language. Additionally, we can report that already an update in DeepSpeech version improved the CER by 3.9% against the earlier studies, which is an important step in the development of Komi ASR. Peer reviewed |
author2 |
The National Library of Finland, Library Network Services Department of Finnish, Finno-Ugrian and Scandinavian Studies |
format |
Conference Object |
author |
Hjortnæs , Nils Partanen, Niko Rießler, Michael Tyers, Francis M. |
author_facet |
Hjortnæs , Nils Partanen, Niko Rießler, Michael Tyers, Francis M. |
author_sort |
Hjortnæs , Nils |
title |
The Relevance of the Source Language in Transfer Learning for ASR |
title_short |
The Relevance of the Source Language in Transfer Learning for ASR |
title_full |
The Relevance of the Source Language in Transfer Learning for ASR |
title_fullStr |
The Relevance of the Source Language in Transfer Learning for ASR |
title_full_unstemmed |
The Relevance of the Source Language in Transfer Learning for ASR |
title_sort |
relevance of the source language in transfer learning for asr |
publishDate |
2022 |
url |
http://hdl.handle.net/10138/351332 |
genre |
Komi language |
genre_facet |
Komi language |
op_relation |
10.33011/computel.v1i.959 Proceedings of the 4th Workshop on the Use of Computational Methods in the Study of Endangered Languages Proceedings of the Workshop on Computational Methods for Endangered Languages 978-1-954085-01-5 Hjortnæs , N , Partanen , N , Rießler , M & Tyers , F M 2021 , The Relevance of the Source Language in Transfer Learning for ASR . in Proceedings of the 4th Workshop on the Use of Computational Methods in the Study of Endangered Languages . vol. 1 , Proceedings of the Workshop on Computational Methods for Endangered Languages , The Association for Computational Linguistics , pp. 63-69 , Workshop on the Use of Computational Methods in the Study of Endangered Languages , 02/03/2021 . https://doi.org/10.33011/computel.v1i.959 conference ORCID: /0000-0001-8584-3880/work/118364555 ORCID: /0000-0002-2397-2860/work/118365886 f93ccf3d-0870-447c-a825-15a0b3524679 http://hdl.handle.net/10138/351332 |
op_rights |
unspecified openAccess info:eu-repo/semantics/openAccess |
container_title |
Proceedings of the Workshop on Computational Methods for Endangered Languages |
container_volume |
1 |
container_issue |
2 |
_version_ |
1787426082927935488 |