Skip to content

Dataset · PerSets

400 hours of Filimo video Persian speech for ASR

PerSets/filimo-persian-asr

Audio recordings in Persian with raw transcriptions for training speech recognition systems.

What's inside

The corpus contains approximately 400 hours of audio extracted from various videos on the Filimo platform in Persian. The text transcripts in the set are raw and have not undergone manual validation, and there is no train-test split; the authors recommend creating one independently.

What you can do with it

  • Training and fine-tuning automatic speech recognition (ASR) models in Persian
  • Pre-training acoustic encoders on conversational speech from movies and series
Tasks
Automatic speech recognition
Languages
Persian
Downloads this month
869
Likes
9

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("PerSets/filimo-persian-asr")