Text-Conditioned Transformer for Automatic Pronunciation Error Detection
Automatic pronunciation error detection (APED) plays an important role in the domain of language learning. As for the previous ASR-based APED methods, the decoded results need to be aligned with the target text so that the errors can be found out. However, since the decoding process and the alignmen...
Main Authors: | , , |
---|---|
Format: | Article in Journal/Newspaper |
Language: | unknown |
Published: |
arXiv
2020
|
Subjects: | |
Online Access: | https://dx.doi.org/10.48550/arxiv.2008.12424 https://arxiv.org/abs/2008.12424 |
id |
ftdatacite:10.48550/arxiv.2008.12424 |
---|---|
record_format |
openpolar |
spelling |
ftdatacite:10.48550/arxiv.2008.12424 2023-05-15T15:06:03+02:00 Text-Conditioned Transformer for Automatic Pronunciation Error Detection Zhang, Zhan Wang, Yuehai Yang, Jianyi 2020 https://dx.doi.org/10.48550/arxiv.2008.12424 https://arxiv.org/abs/2008.12424 unknown arXiv https://dx.doi.org/10.1016/j.specom.2021.04.004 arXiv.org perpetual, non-exclusive license http://arxiv.org/licenses/nonexclusive-distrib/1.0/ Audio and Speech Processing eess.AS FOS Electrical engineering, electronic engineering, information engineering article-journal Article ScholarlyArticle Text 2020 ftdatacite https://doi.org/10.48550/arxiv.2008.12424 https://doi.org/10.1016/j.specom.2021.04.004 2022-03-10T15:55:41Z Automatic pronunciation error detection (APED) plays an important role in the domain of language learning. As for the previous ASR-based APED methods, the decoded results need to be aligned with the target text so that the errors can be found out. However, since the decoding process and the alignment process are independent, the prior knowledge about the target text is not fully utilized. In this paper, we propose to use the target text as an extra condition for the Transformer backbone to handle the APED task. The proposed method can output the error states with consideration of the relationship between the input speech and the target text in a fully end-to-end fashion.Meanwhile, as the prior target text is used as a condition for the decoder input, the Transformer works in a feed-forward manner instead of autoregressive in the inference stage, which can significantly boost the speed in the actual deployment. We set the ASR-based Transformer as the baseline APED model and conduct several experiments on the L2-Arctic dataset. The results demonstrate that our approach can obtain 8.4\% relative improvement on the $F_1$ score metric. : published for Speech Communication journal Article in Journal/Newspaper Arctic DataCite Metadata Store (German National Library of Science and Technology) Arctic Handle The ENVELOPE(161.983,161.983,-78.000,-78.000) |
institution |
Open Polar |
collection |
DataCite Metadata Store (German National Library of Science and Technology) |
op_collection_id |
ftdatacite |
language |
unknown |
topic |
Audio and Speech Processing eess.AS FOS Electrical engineering, electronic engineering, information engineering |
spellingShingle |
Audio and Speech Processing eess.AS FOS Electrical engineering, electronic engineering, information engineering Zhang, Zhan Wang, Yuehai Yang, Jianyi Text-Conditioned Transformer for Automatic Pronunciation Error Detection |
topic_facet |
Audio and Speech Processing eess.AS FOS Electrical engineering, electronic engineering, information engineering |
description |
Automatic pronunciation error detection (APED) plays an important role in the domain of language learning. As for the previous ASR-based APED methods, the decoded results need to be aligned with the target text so that the errors can be found out. However, since the decoding process and the alignment process are independent, the prior knowledge about the target text is not fully utilized. In this paper, we propose to use the target text as an extra condition for the Transformer backbone to handle the APED task. The proposed method can output the error states with consideration of the relationship between the input speech and the target text in a fully end-to-end fashion.Meanwhile, as the prior target text is used as a condition for the decoder input, the Transformer works in a feed-forward manner instead of autoregressive in the inference stage, which can significantly boost the speed in the actual deployment. We set the ASR-based Transformer as the baseline APED model and conduct several experiments on the L2-Arctic dataset. The results demonstrate that our approach can obtain 8.4\% relative improvement on the $F_1$ score metric. : published for Speech Communication journal |
format |
Article in Journal/Newspaper |
author |
Zhang, Zhan Wang, Yuehai Yang, Jianyi |
author_facet |
Zhang, Zhan Wang, Yuehai Yang, Jianyi |
author_sort |
Zhang, Zhan |
title |
Text-Conditioned Transformer for Automatic Pronunciation Error Detection |
title_short |
Text-Conditioned Transformer for Automatic Pronunciation Error Detection |
title_full |
Text-Conditioned Transformer for Automatic Pronunciation Error Detection |
title_fullStr |
Text-Conditioned Transformer for Automatic Pronunciation Error Detection |
title_full_unstemmed |
Text-Conditioned Transformer for Automatic Pronunciation Error Detection |
title_sort |
text-conditioned transformer for automatic pronunciation error detection |
publisher |
arXiv |
publishDate |
2020 |
url |
https://dx.doi.org/10.48550/arxiv.2008.12424 https://arxiv.org/abs/2008.12424 |
long_lat |
ENVELOPE(161.983,161.983,-78.000,-78.000) |
geographic |
Arctic Handle The |
geographic_facet |
Arctic Handle The |
genre |
Arctic |
genre_facet |
Arctic |
op_relation |
https://dx.doi.org/10.1016/j.specom.2021.04.004 |
op_rights |
arXiv.org perpetual, non-exclusive license http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
op_doi |
https://doi.org/10.48550/arxiv.2008.12424 https://doi.org/10.1016/j.specom.2021.04.004 |
_version_ |
1766337719771856896 |