Paper Abstract and Keywords |
Presentation |
2016-10-27 16:00
Word modeling for end-to-end Japanese speech recognition Hitoshi Ito, Aiko Hagiwara, Manon Ichiki, Takeshi Mishima, Shoei Sato (NHK), Akio Kobayashi (NES) SP2016-47 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
In this paper, we propose a novel modeling for end-to-end Japanese speech recognition using Deep Neural Networks(DNN). When we deal with kanji as output layer of DNN, different acoustic features are mapped to the same output labels. This problem is caused by that kanji has multiple readings into a single character, such as On-yomi and Kun-yomi. To resolve the problem, We added words to output layer of DNN, instead of characters.
The words are selected on the basis of the appearance frequency and the rarity of reading in learning data. Our experimental result shows the effectiveness of suppressing a drop of word accuracy in small learning data. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Deep neural networks / Connectionist temporal classification / End-to-end / Acoustic model / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 116, no. 279, SP2016-47, pp. 31-36, Oct. 2016. |
Paper # |
SP2016-47 |
Date of Issue |
2016-10-20 (SP) |
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 |
SP2016-47 |
Conference Information |
Committee |
SP |
Conference Date |
2016-10-27 - 2016-10-27 |
Place (in Japanese) |
(See Japanese page) |
Place (in English) |
Shizuoka University. |
Topics (in Japanese) |
(See Japanese page) |
Topics (in English) |
Learning, Recognition, Synthesis, Dialogue, etc. |
Paper Information |
Registration To |
SP |
Conference Code |
2016-10-SP |
Language |
Japanese |
Title (in Japanese) |
(See Japanese page) |
Sub Title (in Japanese) |
(See Japanese page) |
Title (in English) |
Word modeling for end-to-end Japanese speech recognition |
Sub Title (in English) |
|
Keyword(1) |
Deep neural networks |
Keyword(2) |
Connectionist temporal classification |
Keyword(3) |
End-to-end |
Keyword(4) |
Acoustic model |
Keyword(5) |
|
Keyword(6) |
|
Keyword(7) |
|
Keyword(8) |
|
1st Author's Name |
Hitoshi Ito |
1st Author's Affiliation |
Japan Broadcasting Corporation (NHK) |
2nd Author's Name |
Aiko Hagiwara |
2nd Author's Affiliation |
Japan Broadcasting Corporation (NHK) |
3rd Author's Name |
Manon Ichiki |
3rd Author's Affiliation |
Japan Broadcasting Corporation (NHK) |
4th Author's Name |
Takeshi Mishima |
4th Author's Affiliation |
Japan Broadcasting Corporation (NHK) |
5th Author's Name |
Shoei Sato |
5th Author's Affiliation |
Japan Broadcasting Corporation (NHK) |
6th Author's Name |
Akio Kobayashi |
6th Author's Affiliation |
NHK Engineering System (NES) |
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 |
2016-10-27 16:00:00 |
Presentation Time |
25 minutes |
Registration for |
SP |
Paper # |
SP2016-47 |
Volume (vol) |
vol.116 |
Number (no) |
no.279 |
Page |
pp.31-36 |
#Pages |
6 |
Date of Issue |
2016-10-20 (SP) |
|