Neyshekar

Neyshekar is an open, community-driven Persian speech dataset collected via a web-based crowdsourcing platform at ney.shekar.io. The recordings are provided by a combination of volunteer contributors and paid voice actors, all of whom are native Persian speakers. It supports automatic speech recognition, text-to-speech, speech representation learning, and other downstream Persian speech applications.

Audio is mono, 16-bit PCM, sampled at 16 kHz. Each release represents a stable snapshot of the dataset, enabling reproducible research and consistent benchmarking.

Latest Release - V6.0 (2026-09-01)

• Total samples: 62,279
• Informal samples: 15,222 (24.44%) (identified using Shekar rule-based InformalClassifier)
• Total duration (hours): 99.02
• Average clip duration (seconds): 5.72
• Total tokens: 701,621
• Vocab size: 29,535
• Speakers: 190

Splits

V6.0 is the first release to ship predefined splits. Splits are assigned per speaker, so a single recorder never straddles two sets and the evaluation splits stay speaker-disjoint from train. The speaker_id column makes this checkable: no identifier appears in more than one split.

Split	Samples	Share	Hours	Speakers
train	58,244	93.5%	91.99	134
validation	1,886	3.0%	3.14	26
test	2,149	3.5%	3.88	30

from datasets import load_dataset

dataset = load_dataset("shekar-ai/neyshekar")
train, validation, test = dataset["train"], dataset["validation"], dataset["test"]

Fields

• id: Unique sample identifier
• audio: Speech recording
• text: Persian transcription
• duration: Clip duration in seconds
• speaker_id: Anonymous, stable identifier for the recorder

Links

• GitHub: github.com/amirivojdan/neyshekar
• Zenodo: 10.5281/zenodo.18073632

Terms of Use

Any attempt to identify or uncover the identity of speakers is strictly prohibited.

License

This dataset is released under the CC0 1.0 Universal license. It may be used, modified, and redistributed for any purpose without restriction.