FeruzaSpeech: 60 Hours of Uzbek Read Speech in Dual Scripts

FeruzaSpeech contains 60 hours of high-quality Uzbek read speech recorded by a single native female speaker from Tashkent. Released by k2speech in 2024, the collection provides transcriptions in both Latin and Cyrillic alphabets for academic research. It serves as a resource for low-resource language modeling in the Central Asian region.

Use Cases

• Automatic Speech Recognition (ASR) using the 'audio' and 'text_latin' columns
• Text-to-Speech (TTS) synthesis for the Uzbek language
• Cross-script transliteration modeling between Latin and Cyrillic Uzbek text

Strengths

• 60 hours of audio
• Dual-script transcriptions (Latin and Cyrillic)
• High-quality recordings from a native speaker

Limitations

• Single-speaker bias (female, Tashkent dialect)
• Limited to read speech modality rather than spontaneous conversation
• Small geographic representation restricted to Tashkent

Provenance

Source

k2speech (Arxiv 2410.00035)

Collection Method

Recorded read speech from a single native speaker

Freshness

Last updated July 2025

Geography

Tashkent, Uzbekistan