Investigating the feasibility of harvesting broadcast speech data to develop resources for South African languages

Badenhorst, Jaco; Wet, Febe de

doi:10.55492/dhasa.v3i03.3820

Cited by 2 publications

(6 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Badenhorst & de Wet (2021) it was observed that acoustic match to the harvest data is an important factor for transcription accuracy. The four baseline sub-word systems that were evaluated in this study confirmed this observation.…”

Section: Discussionmentioning

confidence: 99%

“…These indoor speeches were similar to radio A relatively large Afr News test set could be selected from an existing Afr corpus (De Wet et al 2011). In addition, the Afr Messages data, introduced in Badenhorst & de Wet (2021), was included as an example of Afr studio speeches.…”

Section: Test Datamentioning

confidence: 99%

“…Different segmentation techniques were applied depending on the data set. In Badenhorst & de Wet (2021) the segmentation and selection of read speech resources such as radio news bulletins was performed in a two step process.…”

Section: Segmentationmentioning

confidence: 99%

“…The data harvesting initiative is supported by the South African Centre for Digital Language Resources (SADiLaR[1]) as part of their efforts to develop indigenous language resources. A feasibility study focusing on the collection, building and testing of automatic transcription systems yielded initial transcriptions of broadcast speech data in two languages (Badenhorst & de Wet 2021). The results of the study indicated that the proposed approach could also be used to collect data in other languages.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Gauging the accuracy of automatic speech data harvesting in five under-resourced languages

Badenhorst¹,

Wet

2022

JDHASA

View full text Add to dashboard Cite

Recent research on deep-learning architectures has resulted in substantial improvements in automatic speech recognition accuracy. The leaps of progress made in well-resourced languages can be attributed to the fact that these architectures are able to effectively represent spoken language in all its diversity and complexity. However, developing advanced models of a language without appropriate corpora of speech and text data remains a challenge. For many under-resourced languages, including those spoken in South Africa, such resources simply do not exist. The aim of the work reported on in this paper is to address this situation by investigating the possibility to create diverse speech resources from unannotated broadcast data. The paper describes how existing speech and text resources were used to develop a semi-automatic data harvesting procedure for two genres of broadcast data, namely news bulletins and radio dramas. It was found that adapting acoustic models with less than 10 hours of manually annotated data from the same domain significantly reduced transcription error rates for speaking styles and acoustic conditions that are not represented in any of the existing speech corpora. Results also indicated that much more automatically transcribed adaptation data is required to achieve similar results.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Test Datamentioning

confidence: 99%

Section: Segmentationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Gauging the accuracy of automatic speech data harvesting in five under-resourced languages

Badenhorst¹,

Wet

2022

JDHASA

View full text Add to dashboard Cite

show abstract

“…The feasibility of harvesting radio broadcast speech data for the development of ASR systems for South African languages was first evaluated in Badenhorst & de Wet (2021). In this work, a semiautomatic data harvesting procedure was proposed.…”

Section: Previous Workmentioning

confidence: 99%

Exploring ASR fine-tuning on limited domain-specific data for low-resource languages

Mak,

Govender,

Badenhorst

2024

JDHASA

View full text Add to dashboard Cite

The majority of South Africa's eleven languages are low-resourced, posing a major challenge to Automatic Speech Recognition (ASR) development. Modern ASR systems require an extensive amount of data that is extremely difficult to find for lowresourced languages. In addition, available speech and text corpora for these languages predominantly revolve around government, political and biblical content. Consequently, this hinders the ability of ASR systems developed for these languages to perform well especially when evaluating data outside of these domains. To alleviate this problem, the Icefall Kaldi II toolkit introduced new transformer model scripts, facilitating the adaptation of pre-trained models using limited adaptation data. In this paper, we explored the technique of using pre-trained ASR models in a domain where more data is available (government data) and adapted it to an entirely different domain with limited data (broadcast news data). The objective was to assess whether such techniques can surpass the accuracy of prior ASR models developed for these lan-guages. Our results showed that the Conformer connectionist temporal classification (CTC) model obtained lower word error rates by a large margin in comparison to previous TDNN-F models evaluated on the same datasets. This research signifies a step forward in mitigating data scarcity challenges and enhancing ASR performance for low-resourced languages in South Africa.

show abstract

Investigating the feasibility of harvesting broadcast speech data to develop resources for South African languages

Cited by 2 publications

References 0 publications

Gauging the accuracy of automatic speech data harvesting in five under-resourced languages

Gauging the accuracy of automatic speech data harvesting in five under-resourced languages

Exploring ASR fine-tuning on limited domain-specific data for low-resource languages

Contact Info

Product

Resources

About