README.md

metadata

dataset_info:
  features:
    - name: file_name
      dtype: string
    - name: uni
      dtype: string
    - name: wylie
      dtype: string
    - name: url
      dtype: string
    - name: dept
      dtype: string
    - name: grade
      dtype: int64
    - name: char_len
      dtype: int64
    - name: audio_len
      dtype: float64
    - name: original_id
      dtype: string
    - name: strata
      dtype: string
    - name: __index_level_0__
      dtype: int64
  splits:
    - name: train
      num_bytes: 4017714
      num_examples: 6988
    - name: validation
      num_bytes: 212173
      num_examples: 368
  download_size: 1451597
  dataset_size: 4229887
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*

Dataset Structure

• Features:

  • file_name: Name of the file.
  • uni: Tibetan Unicode text.
  • wylie: Wylie transliteration.
  • url: Source URL.
  • dept: Department or category.
  • grade: Grade/class level.
  • char_len: Number of characters.
  • audio_len: Length of audio (seconds).
  • original_id: Original document or audio ID.
  • strata: Data strata or group.
  • __index_level_0__: Index.

• Splits: train, validation

---

📊 Split-wise Statistics

Split	# Samples	Total Char Length	Total Audio Length (hr)
train	6988	613604	10.5
validation	368	32462	0.56
Total	7356	646066	11.06

strata value counts:

strata	train_count	validation_count
70-80__long__Teaching	3064	161
70-80__medium__Teaching	1347	71
80-90__long__Practice	1011	53
70-80__long__Q&A	635	34
70-80__long__Prayer	593	31
70-80__medium__Prayer	172	9
70-80__short__Teaching	105	6
70-80__medium__Q&A	61	3

🚀 Usage

from datasets import load_dataset
ds = load_dataset("your_namespace/your_dataset", split="train")