| Content Provider |
IEEE Xplore Digital Library |
| Author |
Cui, Jia Kingsbury, Brian Ramabhadran, Bhuvana Sethy, Abhinav Audhkhasi, Kartik Cui, Xiaodong Kislal, Ellen Mangu, Lidia Nussbaum-Thom, Markus Picheny, Michael Tuske, Zoltan Golik, Pavel Schluter, Ralf Ney, Hermann Gales, Mark J. F. Knill, Kate M. Ragni, Anton Wang, Haipeng Woodland, Phil |
| Copyright Year |
2015 |
| Description |
Author affiliation: Computer Science Department, RWTH Aachen University, 52056 Aachen, Germany (Tuske, Zoltan; Golik, Pavel; Schluter, Ralf; Ney, Hermann) || IBM Watson, 1101 Kitchawan Rd, Yorktown Heights, NY, 10598, U.S.A. (Cui, Jia; Kingsbury, Brian; Ramabhadran, Bhuvana; Sethy, Abhinav; Audhkhasi, Kartik; Cui, Xiaodong; Kislal, Ellen; Mangu, Lidia; Nussbaum-Thom, Markus; Picheny, Michael) || Cambridge University Engineering Department, Trumpington Street, Cambridge, CB2 1PZ, UK (Gales, Mark J. F.; Knill, Kate M.; Ragni, Anton; Wang, Haipeng; Woodland, Phil) |
| Abstract |
This paper examines the impact of multilingual (ML) acoustic representations on Automatic Speech Recognition (ASR) and keyword search (KWS) for low resource languages in the context of the OpenKWS15 evaluation of the IARPA Babel program. The task is to develop Swahili ASR and KWS systems within two weeks using as little as 3 hours of transcribed data. Multilingual acoustic representations proved to be crucial for building these systems under strict time constraints. The paper discusses several key insights on how these representations are derived and used. First, we present a data sampling strategy that can speed up the training of multilingual representations without appreciable loss in ASR performance. Second, we show that fusion of diverse multilingual representations developed at different LORELEI sites yields substantial ASR and KWS gains. Speaker adaptation and data augmentation of these representations improves both ASR and KWS performance (up to 8.7% relative). Third, incorporating un-transcribed data through semi-supervised learning, improves WER and KWS performance. Finally, we show that these multilingual representations significantly improve ASR and KWS performance (relative 9% for WER and 5% for MTWV) even when forty hours of transcribed audio in the target language is available. Multilingual representations significantly contributed to the LORELEI KWS systems winning the OpenKWS15 evaluation. |
| Starting Page |
259 |
| Ending Page |
266 |
| File Size |
231518 |
| Page Count |
8 |
| File Format |
PDF |
| e-ISBN |
9781479972913 |
| DOI |
10.1109/ASRU.2015.7404803 |
| Language |
English |
| Publisher |
Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Publisher Date |
2015-12-13 |
| Publisher Place |
USA |
| Access Restriction |
Subscribed |
| Rights Holder |
Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Subject Keyword |
Training Training data Keyword search Context Data models Acoustics Neural networks BABEL Multilingual Representation Hierarchical Deep Neural Network Keyword Search |
| Content Type |
Text |
| Resource Type |
Article |
