| 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 | 2017-03-07 14:00 CTC network with explicit representation vector of Markov property Yuta Kawachi, Taichi Asami, Yoshikazu Yamaguchi, Yushi Aono (NTT)  IBISML2016-111 |
|
| Abstract | (in Japanese) | (See Japanese page) |
| (in English) | Current neural acoustic models are incapable of utilizing language resources except speech transcriptions. So toward the realization of true neural end-to-end speech recognition, this property is problematic. In this paper, we present a method for general sequence classification task based on connectionist temporal classification with a structure keeping the previously output symbols and a matrix explicitly represents bigram properties. In our experiment, we defined a problem easily solved by using Markov properties of output sequences. Finally we could obtain interpretable neural network parameters and accuracy improvements from the baseline LSTM-CTC model. | |
| Keyword | (in Japanese) | (See Japanese page) |
| (in English) | temporal classification / deep learning / CTC / end-to-end / / / / | |
| Reference Info. | IEICE Tech. Rep., vol. 116, no. 500, IBISML2016-111, pp. 83-88, March 2017. | |
| Paper # | IBISML2016-111 | |
| Date of Issue | 2017-02-27 (IBISML) | |
| ISSN | Print edition: ISSN 0913-5685 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 | IBISML2016-111 |
|
| Conference Information | |
| Committee | IBISML |
| Conference Date | 2017-03-06 - 2017-03-07 |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Tokyo Institute of Technology |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Statistical Mathematics, Machine Learning, Data Mining, etc. |
| Paper Information | |
| Registration To | IBISML |
| Conference Code | 2017-03-IBISML |
| Language | Japanese |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | CTC network with explicit representation vector of Markov property |
| Sub Title (in English) | |
| Keyword(1) | temporal classification |
| Keyword(2) | deep learning |
| Keyword(3) | CTC |
| Keyword(4) | end-to-end |
| Keyword(5) | |
| Keyword(6) | |
| Keyword(7) | |
| Keyword(8) | |
| 1st Author's Name | Yuta Kawachi |
| 1st Author's Affiliation | NTT MedNTT Corporationia Intelligence Laboratories (NTT) |
| 2nd Author's Name | Taichi Asami |
| 2nd Author's Affiliation | NTT MedNTT Corporationia Intelligence Laboratories (NTT) |
| 3rd Author's Name | Yoshikazu Yamaguchi |
| 3rd Author's Affiliation | NTT MediaNTT Corporation Intelligence Laboratories (NTT) |
| 4th Author's Name | Yushi Aono |
| 4th Author's Affiliation | NTT Corporation (NTT) |
| Speaker | Author-1 |
| Date Time | 2017-03-07 14:00:00 |
| Presentation Time | 30 minutes |
| Registration for | IBISML |
| Paper # | IBISML2016-111 |
| Volume (vol) | vol.116 |
| Number (no) | no.500 |
| Page | pp.83-88 |
| #Pages | 6 |
| Date of Issue | 2017-02-27 (IBISML) |
[Return to Top Page]