main_dataset / README.md
Elyordev's picture
Update README.md
c6abd6b verified
|
raw
history blame
1.55 kB
metadata
language:
  - uz
tags:
  - automatic-speech-recognition
  - audio
  - uzbek
task_categories:
  - automatic-speech-recognition
task_ids:
  - automatic-speech-recognition
license: cc-by-4.0
datasets:
  - custom
pretty_name: Uzbek Speech-to-Text Dataset

Uzbek Speech-to-Text Dataset

Maqsad

Ushbu dataset O‘zbekiston tilidagi nutqni matnga aylantirish (Speech-to-Text) bo‘yicha modellarni o‘qitish va baholash uchun yaratilgan. Bu dataset avtomatik nutqni aniqlash (ASR) texnologiyalarini rivojlantirish maqsadida ishlab chiqilgan.

Strukturasi

Dataset quyidagi tuzilishga ega:

  • audio/: Audio fayllar saqlanadigan papka.
  • meta.csv: Audio fayllar va ularga mos transkripsiyalarni bog‘lovchi meta fayl.
  • Transkripsiyalar: Har bir audio fayl uchun to‘g‘ri yozilgan matn.

meta.csv Fayl Tuzilishi

id path sentence
1 audio/audio_0001/00000001.wav Bolalar maydonchalari qurilgan.
2 audio/audio_0002/00000002.wav Ro‘yxatni Matiz boshqarmoqda.

Foydalanish

Datasetni yuklab olish va ishlatish uchun quyidagi kodni ishlating:

from datasets import load_dataset

# Datasetni yuklash
dataset = load_dataset("username/dataset-name")

# Train qismidagi birinchi yozuvni ko‘rsatish
print(dataset["train"][0])

# Audio faylni o‘qish va tinglash
audio = dataset["train"][0]["audio"]["array"]
print("Audio:", audio)