FLEURS-Kobani: Northern Kurdish Speech Benchmark with 18 Hours of Audio

FLEURS-Kobani is a speech dataset for Northern Kurdish (Kurmanji, ISO 639-3: KMR), designed as an extension of the FLEURS benchmark. It contains 5,162 utterances totaling 18 hours and 24 minutes of audio from 31 native speakers. The dataset was created by author 'aranemini' and was last updated on Hugging Face in April 2026.

Use Cases

• Train Automatic Speech Recognition (ASR) models based on the described Northern Kurdish speech data.
• Develop Speech-to-Text Translation (S2TT) systems for Northern Kurdish.
• Research Speech-to-Speech Translation (S2ST) using the provided audio utterances.
• Benchmark multilingual speech model performance on the Kurmanji language.

Strengths

• Contains 5,162 distinct speech utterances.
• Provides 18 hours and 24 minutes of total audio duration.
• Includes recordings from 31 native speakers, with 26 female and 5 male participants.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.
• The dataset's geographic origin and recording conditions are not specified.

Provenance

Source

huggingface

Collection Method

Likely recorded speech from native speakers, as an extension of the FLEURS benchmark.

Freshness

Last updated 2026-04-20 20:45:56; freshness should be verified.