Publication Details
Multilingual Region-Dependent Transforms
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Grézl František, Ing., Ph.D. (DCGM)
Veselý Karel, Ing., Ph.D. (DCGM)
Černocký Jan, prof. Dr. Ing. (DCGM)
Automatic speech recognition, Region-Dependent Transforms, Multilingual speech recognition, Feedforward neural networks
This paper presented our further steps in the development of a feature extraction scheme easily transferable to a new language with severely limited training data.
In recent years, trained feature extraction (FE) schemes based on neural networks have replaced or complemented traditional approaches in top performing systems. This paper deals with FE in multilingual scenarios with a target language with low amount of transcribed data. Continuing our previous work on multilingual training of Stacked Bottle-Neck Neural Network FE schemes, we concentrate on improving the discriminatively trained Region- Dependent Transforms. We show that multilingual training of RDT can be implemented by merging statistics from several languages. In our case we used up to 11 source languages to build a FE which generalize well for a new language. This allows us to build a strong bootstrapping model for the final ASR system. The results are produced on IARPA Babel data.
@inproceedings{BUT130965,
author="Martin {Karafiát} and Lukáš {Burget} and František {Grézl} and Karel {Veselý} and Jan {Černocký}",
title="Multilingual Region-Dependent Transforms",
booktitle="Proceedings of the 41th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2016), 2016",
year="2016",
pages="5430--5434",
publisher="IEEE Signal Processing Society",
address="Shanghai",
doi="10.1109/ICASSP.2016.7472715",
isbn="978-1-4799-9988-0",
url="https://www.fit.vut.cz/research/publication/11146/"
}