Rohingya Language Broadcast Audio for ASR, Segmented from VOA News

The first public Rohingya language ASR dataset in AI history contains broadcast audio recordings from the Voice of America Rohingya Service. Each file represents a daily news segment, typically 30 minutes long, automatically segmented into 5–15 second chunks. The dataset was created by author 'freococo' and was last updated on June 14, 2025.

Use Cases

• Self-supervised pretraining for speech models based on broadcast audio chunks.
• Automatic speech recognition model development for the Rohingya language.
• Language identification tasks based on Rohingya audio segments.
• Research on low-resource language processing using news broadcast content.

Strengths

• First public ASR dataset for the Rohingya language, addressing a significant gap.
• Audio is segmented into consistent 5–15 second chunks suitable for model training.
• Source material is from the Voice of America Rohingya Service, providing authentic broadcast content.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.
• Data may reflect temporal and editorial bias inherent to its single source, VOA broadcasts.

Provenance

Source

Voice of America (VOA) Rohingya Service broadcast recordings.

Collection Method

Daily news segments were automatically segmented into audio chunks.

Freshness

Last updated 2025-06-14 06:30:24