VLSP2020 VinAI 100H: 100 Hours of Vietnamese Speech for ASR

Vietnamese speech data comprising 100 hours of audio, released for the VLSP 2020 Automatic Speech Recognition challenge by VinBigData. The dataset is an unofficial mirror hosted on Hugging Face by the user 'doof-ferb', with the original source linked to a 2020 community-sharing event from the VinBigData Institute.

Use Cases

• Training Vietnamese speech recognition models based on the 100 hours of audio data.
• Benchmarking ASR system performance for the Vietnamese language based on the challenge dataset.
• Fine-tuning pre-trained multilingual speech models on Vietnamese audio data.
• Conducting linguistic or acoustic analysis of Vietnamese speech patterns.

Strengths

• Contains 100 hours of Vietnamese speech audio, providing a substantial resource for a low-resource language.
• Released for a formal ASR challenge (VLSP 2020), suggesting a benchmark standard.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

VinBigData Institute, released for the VLSP 2020 workshop.

Collection Method

Likely collected and curated for a public ASR challenge.

Time Range

Associated with the 2020 VLSP workshop.

Freshness

Last updated 2025-04-20 20:19:38; freshness should be verified.

Geography

Vietnam (Vietnamese language data).