Faculty of Information Technology, BUT

Publication Details

Data selection by sequence summarizing neural network in mismatch condition training

ŽMOLÍKOVÁ Kateřina, KARAFIÁT Martin, VESELÝ Karel, DELCROIX Marc, WATANABE Shinji, BURGET Lukáš and ČERNOCKÝ Jan. Data selection by sequence summarizing neural network in mismatch condition training. In: Proceedings of Interspeech 2016. San Francisco: International Speech Communication Association, 2016, pp. 2354-2358. ISBN 978-1-5108-3313-5. Available from: https://www.semanticscholar.org/paper/Data-Selection-by-Sequence-Summarizing-Neural-Zmol%C3%ADkov%C3%A1-Karafi%C3%A1t/bc1832e8b8d4e5edf987e1562b578bd9aa5e18a9
Czech title
Výběr dat pomocí sekvenční sumarizační neuronové sítě v trénování na datech z odlišných podmínek
Type
conference paper
Language
english
Authors
Žmolíková Kateřina, Ing. (DCGM FIT BUT)
Karafiát Martin, Ing., Ph.D. (DCGM FIT BUT)
Veselý Karel, Ing., Ph.D. (DCGM FIT BUT)
Delcroix Marc (NTT)
Watanabe Shinji, Dr. (JHU)
Burget Lukáš, doc. Ing., Ph.D. (DCGM FIT BUT)
Černocký Jan, doc. Dr. Ing. (DCGM FIT BUT)
URL
Keywords
Automatic speech recognition, Data augmentation, Data selection, Mismatch training condition, Sequence summarization
Abstract
Data augmentation is a simple and efficient technique to improve the robustness of a speech recognizer when deployed in mismatched training-test conditions. Our paper proposes a new approach for selecting data with respect to similarity of acoustic conditions. The similarity is computed based on a sequence summarizing neural network which extracts vectors containing acoustic summary (e.g. noise and reverberation characteristics) of an utterance. Several configurations of this network and different methods of selecting data using these "summary-vectors" were explored. The results are reported on a mismatched condition using AMI training set with the proposed data selection and CHiME3 test set.
Annotation
Data augmentation is a simple and efficient technique to improve the robustness of a speech recognizer when deployed in mismatched training-test conditions. Our paper proposes a new approach for selecting data with respect to similarity of acoustic conditions. The similarity is computed based on a sequence summarizing neural network which extracts vectors containing acoustic summary (e.g. noise and reverberation characteristics) of an utterance. Several configurations of this network and different methods of selecting data using these "summary-vectors" were explored. The results are reported on a mismatched condition using AMI training set with the proposed data selection and CHiME3 test set.
Published
2016
Pages
2354-2358
Proceedings
Proceedings of Interspeech 2016
Conference
Interspeech 2016, San Francisco, US
ISBN
978-1-5108-3313-5
Publisher
International Speech Communication Association
Place
San Francisco, US
DOI
BibTeX
@INPROCEEDINGS{FITPUB11271,
   author = "Kate\v{r}ina \v{Z}mol\'{i}kov\'{a} and Martin Karafi\'{a}t and Karel Vesel\'{y} and Marc Delcroix and Shinji Watanabe and Luk\'{a}\v{s} Burget and Jan \v{C}ernock\'{y}",
   title = "Data selection by sequence summarizing neural network in mismatch condition training",
   pages = "2354--2358",
   booktitle = "Proceedings of Interspeech 2016",
   year = 2016,
   location = "San Francisco, US",
   publisher = "International Speech Communication Association",
   ISBN = "978-1-5108-3313-5",
   doi = "10.21437/Interspeech.2016-741",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/11271"
}
Back to top