Shemo Diarization: 50 Hours of Synthetic Multi-Speaker Persian Speech

Persian (Farsi) synthetic multi-speaker speech dataset for speaker diarization. It contains approximately 50 hours of audio across 5,000 tracks, built from utterances in the Shemo dataset and processed through a synthesis framework. The dataset was created by atiyehghm and last updated on Hugging Face in May 2026.

Use Cases

• Training speaker diarization models based on simulated conversational dynamics.
• Evaluating diarization system performance on Persian-language audio.
• Benchmarking synthetic speech generation frameworks for multi-speaker scenarios.
• Developing speech processing pipelines for Persian conversational analysis.

Strengths

• Approximately 50 hours of total audio duration provides a substantial resource.
• Contains 5,000 individual audio tracks for model training and evaluation.
• Specifically designed for the Persian (Farsi) language, addressing a less common language domain.
• Uses a synthesis framework to simulate realistic multi-speaker dynamics.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is synthetic, which may not fully capture the acoustic variability of real-world recordings.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Built using utterances from the Shemo dataset.

Collection Method

Processed through a synthesis framework to simulate conversational dynamics.

Freshness

Last updated 2026-05-12 06:09:03; freshness should be verified.