Information: Join today and make your research activities more affordable! Technical workshop participation fees and annual registration fees are available at member rates.
Notice: [Important] Announcement of Changes to Registration Fee Payment and Manuscript Upload Procedures for IEICE Technical Meetings
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 2020-12-02 09:40
Fast End-to-End Speech Recognition with CTC and Mask Predict
Yosuke Higuchi (Waseda Univ.), Hirofumi Inaguma (Kyoto Univ.), Shinji Watanabe (JHU), Tetsuji Ogawa, Tetsunori Kobayashi (Waseda Univ.) NLC2020-13 SP2020-16
Abstract (in Japanese) (See Japanese page) 
(in English) We present a fast non-autoregressive (NAR) end-to-end automatic speech recognition (E2E-ASR) framework, which generates a sequence by refining outputs of the connectionist temporal classification (CTC) via mask prediction. Many of the previous studies on E2E-ASR focus on an textit{autoregressive} (AR) model: each output token is generated by conditioning on previously generated tokens, at the cost of requiring as many iterations as the output length. On the other hand, NAR models can simultaneously generate tokens within a constant number of iterations, which results in significant inference time reduction and better suits end-to-end ASR model for real-world scenarios. In this work, we train an E2E-ASR model with joint objectives of CTC and mask prediction. During inference, the greedy CTC output is refined by mask prediction, where errors in the CTC output are recovered by taking account of conditional dependence between output tokens. Experimental results on different speech recognition tasks show that the proposed model achieves fast inference time ($<$0.1 RTF using CPU), outperforming a standard CTC model and achieving competitive results to the AR models.
Keyword (in Japanese) (See Japanese page) 
(in English) end-to-end speech recognition / connectionist temporal classification / non-autoregressive sequence generation / / / / /  
Reference Info. IEICE Tech. Rep., vol. 120, no. 271, SP2020-16, pp. 1-6, Dec. 2020.
Paper # SP2020-16 
Date of Issue 2020-11-25 (NLC, 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 NLC2020-13 SP2020-16

Conference Information
Committee NLC IPSJ-NL SP IPSJ-SLP  
Conference Date 2020-12-02 - 2020-12-03 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To SP 
Conference Code 2020-12-NLC-NL-SP-SLP 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Fast End-to-End Speech Recognition with CTC and Mask Predict 
Sub Title (in English)  
Keyword(1) end-to-end speech recognition  
Keyword(2) connectionist temporal classification  
Keyword(3) non-autoregressive sequence generation  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Yosuke Higuchi  
1st Author's Affiliation Waseda University (Waseda Univ.)
2nd Author's Name Hirofumi Inaguma  
2nd Author's Affiliation Kyoto Univeristy (Kyoto Univ.)
3rd Author's Name Shinji Watanabe  
3rd Author's Affiliation Johns Hopkins University (JHU)
4th Author's Name Tetsuji Ogawa  
4th Author's Affiliation Waseda University (Waseda Univ.)
5th Author's Name Tetsunori Kobayashi  
5th Author's Affiliation Waseda University (Waseda Univ.)
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 ()
21st Author's Name  
21st Author's Affiliation ()
22nd Author's Name  
22nd Author's Affiliation ()
23rd Author's Name  
23rd Author's Affiliation ()
24th Author's Name  
24th Author's Affiliation ()
25th Author's Name  
25th Author's Affiliation ()
26th Author's Name / /
26th Author's Affiliation ()
()
27th Author's Name / /
27th Author's Affiliation ()
()
28th Author's Name / /
28th Author's Affiliation ()
()
29th Author's Name / /
29th Author's Affiliation ()
()
30th Author's Name / /
30th Author's Affiliation ()
()
31st Author's Name / /
31st Author's Affiliation ()
()
32nd Author's Name / /
32nd Author's Affiliation ()
()
33rd Author's Name / /
33rd Author's Affiliation ()
()
34th Author's Name / /
34th Author's Affiliation ()
()
35th Author's Name / /
35th Author's Affiliation ()
()
36th Author's Name / /
36th Author's Affiliation ()
()
Speaker Author-1 
Date Time 2020-12-02 09:40:00 
Presentation Time 20 minutes 
Registration for SP 
Paper # NLC2020-13, SP2020-16 
Volume (vol) vol.120 
Number (no) no.270(NLC), no.271(SP) 
Page pp.1-6 
#Pages 6 
Date of Issue 2020-11-25 (NLC, SP) 

[Return to Top Page]

[Return to IEICE Web Page]
The Institute of Electronics, Information and Communication Engineers (IEICE), Japan