IEICE Technical Committee Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2023-03-01 10:10
Vocabulary-Set Decomposition and Multi-task Learning for Target Vocabulary Extraction in Japanese Speech Recognition
Aoi Ito (LINE/Hosei Univ.), Tatsuya Komatsu, Yusuke Fujita (LINE) EA2022-102 SIP2022-146 SP2022-66
Abstract (in Japanese) (See Japanese page) 
(in English) This paper proposes a target vocabulary extraction method for Japanese speech recognition models based on vocabulary set decomposition and multi-task learning. In the proposed method, when a certain vocabulary set, e.g., a series consisting only of katakana, is to be extracted, multi-task learning is performed by calculating the losses for the target katakana sequence and the non-katakana sequence. By combining the losses for the complementary sequence, the target vocabulary can be extracted with higher accuracy than in single-task learning using only the target vocabulary. Experiments show that the proposed method can extract the target vocabulary more accurately than the target vocabulary filtering and the single-task learning method using only the target vocabulary.
Keyword (in Japanese) (See Japanese page) 
(in English) Automatic speech recognition / CTC / Self-Conditioned CTC / target vocabulary extraction / / / /  
Reference Info. IEICE Tech. Rep., vol. 122, no. 389, SP2022-66, pp. 159-164, Feb. 2023.
Paper # SP2022-66 
Date of Issue 2023-02-21 (EA, SIP, SP) 
ISSN Online edition: ISSN 2432-6380
Copyright
and
reproduction
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034)
Download PDF EA2022-102 SIP2022-146 SP2022-66

Conference Information
Committee SP IPSJ-SLP EA SIP  
Conference Date 2023-02-28 - 2023-03-01 
Place (in Japanese) (See Japanese page) 
Place (in English)  
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To SP 
Conference Code 2023-02-SP-SLP-EA-SIP 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Vocabulary-Set Decomposition and Multi-task Learning for Target Vocabulary Extraction in Japanese Speech Recognition 
Sub Title (in English)  
Keyword(1) Automatic speech recognition  
Keyword(2) CTC  
Keyword(3) Self-Conditioned CTC  
Keyword(4) target vocabulary extraction  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Aoi Ito  
1st Author's Affiliation LINE Corporation/Hosei University (LINE/Hosei Univ.)
2nd Author's Name Tatsuya Komatsu  
2nd Author's Affiliation LINE Corporation (LINE)
3rd Author's Name Yusuke Fujita  
3rd Author's Affiliation LINE Corporation (LINE)
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
Speaker Author-1 
Date Time 2023-03-01 10:10:00 
Presentation Time 20 minutes 
Registration for SP 
Paper # EA2022-102, SIP2022-146, SP2022-66 
Volume (vol) vol.122 
Number (no) no.387(EA), no.388(SIP), no.389(SP) 
Page pp.159-164 
#Pages
Date of Issue 2023-02-21 (EA, SIP, SP) 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan