Bangor Transcription Bank: 60 Hours of Natural Welsh Speech

60 hours of natural speech segments from over fifty contributors are provided as MP3 files with corresponding verbatim transcriptions in TSV format. The majority of the speech is described as spontaneous and natural. This collection is shared under an open CC0 license by DewiBrynJones and was last updated on Hugging Face in May 2026.

Use Cases

• Training automatic speech recognition (ASR) models based on natural Welsh speech audio.
• Fine-tuning speech-to-text systems based on verbatim transcriptions.
• Analyzing spontaneous speech patterns and acoustic features in the Welsh language.

Strengths

• Contains 60 hours of audio data.
• Includes verbatim transcriptions aligned with audio segments.
• Focuses on natural, spontaneous speech from over fifty contributors.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total file size are unknown, which may limit suitability assessment.

Provenance

Source

DewiBrynJones on Hugging Face.

Collection Method

Likely collected from voluntary contributors.

Freshness

Last updated 2026-05-21 11:57:13; freshness should be verified.

Geography

Likely focused on Welsh language speakers.