ResearchSpace

The usefulness of imperfect speech data for ASR development in low-resource languages

Show simple item record

dc.contributor.author Badenhorst, Jacob AC
dc.contributor.author De Wet, Febe
dc.date.accessioned 2019-10-28T10:19:19Z
dc.date.available 2019-10-28T10:19:19Z
dc.date.issued 2019-08
dc.identifier.citation Badenhorst, J.A.C. & De Wet, F. 2019. The usefulness of imperfect speech data for ASR development in low-resource languages. Information, Vol. 10, no. 9, pp. 1-6 en_US
dc.identifier.issn 2078-2489
dc.identifier.uri https://www.mdpi.com/2078-2489/10/9/268
dc.identifier.uri Doi:10.3390/info10090268
dc.identifier.uri http://hdl.handle.net/10204/11196
dc.description © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). en_US
dc.description.abstract When the National Centre for Human Language Technology (NCHLT) Speech corpus was released, it created various opportunities for speech technology development in the 11 official, but critically under-resourced, languages of South Africa. Since then, the substantial improvements in acoustic modeling that deep architectures achieved for well-resourced languages ushered in a new data requirement: their development requires hundreds of hours of speech. A suitable strategy for the enlargement of speech resources for the South African languages is therefore required. The first possibility was to look for data that has already been collected but has not been included in an existing corpus. Additional data was collected during the NCHLT project that was not included in the official corpus: it only contains a curated, but limited subset of the data. In this paper, we first analyze the additional resources that could be harvested from the auxiliary NCHLT data. We also measure the effect of this data on acoustic modeling. The analysis incorporates recent factorized time-delay neural networks (TDNN-F). These models significantly reduce phone error rates for all languages. In addition, data augmentation and cross-corpus validation experiments for a number of the datasets illustrate the utility of the auxiliary NCHLT data. en_US
dc.language.iso en en_US
dc.publisher MDPI en_US
dc.relation.ispartofseries Workflow;22685
dc.subject Automatic speech recognition en_US
dc.subject Kaldi en_US
dc.subject Low-resource languages en_US
dc.subject Speech data en_US
dc.subject Speech technology en_US
dc.subject Time-delay neural networks en_US
dc.title The usefulness of imperfect speech data for ASR development in low-resource languages en_US
dc.type Article
dc.identifier.apacitation Badenhorst, J. A., & De Wet, F. (2019). The usefulness of imperfect speech data for ASR development in low-resource languages. http://hdl.handle.net/10204/11196 en_ZA
dc.identifier.chicagocitation Badenhorst, Jacob AC, and Febe De Wet "The usefulness of imperfect speech data for ASR development in low-resource languages." (2019) http://hdl.handle.net/10204/11196 en_ZA
dc.identifier.vancouvercitation Badenhorst JA, De Wet F. The usefulness of imperfect speech data for ASR development in low-resource languages. 2019; http://hdl.handle.net/10204/11196. en_ZA
dc.identifier.ris TY - Article AU - Badenhorst, Jacob AC AU - De Wet, Febe AB - When the National Centre for Human Language Technology (NCHLT) Speech corpus was released, it created various opportunities for speech technology development in the 11 official, but critically under-resourced, languages of South Africa. Since then, the substantial improvements in acoustic modeling that deep architectures achieved for well-resourced languages ushered in a new data requirement: their development requires hundreds of hours of speech. A suitable strategy for the enlargement of speech resources for the South African languages is therefore required. The first possibility was to look for data that has already been collected but has not been included in an existing corpus. Additional data was collected during the NCHLT project that was not included in the official corpus: it only contains a curated, but limited subset of the data. In this paper, we first analyze the additional resources that could be harvested from the auxiliary NCHLT data. We also measure the effect of this data on acoustic modeling. The analysis incorporates recent factorized time-delay neural networks (TDNN-F). These models significantly reduce phone error rates for all languages. In addition, data augmentation and cross-corpus validation experiments for a number of the datasets illustrate the utility of the auxiliary NCHLT data. DA - 2019-08 DB - ResearchSpace DP - CSIR KW - Automatic speech recognition KW - Kaldi KW - Low-resource languages KW - Speech data KW - Speech technology KW - Time-delay neural networks LK - https://researchspace.csir.co.za PY - 2019 SM - 2078-2489 T1 - The usefulness of imperfect speech data for ASR development in low-resource languages TI - The usefulness of imperfect speech data for ASR development in low-resource languages UR - http://hdl.handle.net/10204/11196 ER - en_ZA


Files in this item

This item appears in the following Collection(s)

Show simple item record