README.md

metadata

license: mit
dataset_info:
  features:
    - name: audio
      dtype: string
    - name: file_name
      dtype: string
    - name: folder
      dtype: string
    - name: transcription
      dtype: string
    - name: confidence
      dtype: string
    - name: trim_start_seconds
      dtype: float64
    - name: trim_end_seconds
      dtype: float64
    - name: reason_start
      dtype: string
    - name: reason_end
      dtype: string
    - name: has_incomplete_words
      dtype: bool
    - name: notes
      dtype: string
    - name: original_path
      dtype: string
    - name: batch_index
      dtype: int64
    - name: global_index
      dtype: int64
  splits:
    - name: train
      num_bytes: 12442238
      num_examples: 24369
  download_size: 3223885
  dataset_size: 12442238
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Sam04/BlueLionEcho_tran

This dataset contains transcribed audio files organized in folders for scalability.

Dataset Structure

The dataset is organized with:

• Audio files: Stored in audio_XXXXX/ folders (5000 files per folder)
• Metadata: Stored in data_XXXXX/ folders as parquet files

This organization follows Hugging Face best practices for datasets with millions of files.

Statistics

• Total files: 6,456
• Total batches: 2525
• Audio folders: 3
• Files per folder: max 5000

Loading the Dataset

from datasets import load_dataset

# Load the complete dataset
dataset = load_dataset("Sam04/BlueLionEcho_tran")

# The 'audio' column contains paths like "audio_00000/0000000001_filename.wav"
# Files are automatically resolved when accessing the dataset

Folder Organization

Audio files are distributed across folders to respect HuggingFace storage limits:

• audio_00000/: Files 0-4,999
• audio_00001/: Files 5,000-9,999
• etc.

Metadata (parquet files) are grouped by batch ranges:

• data_00000/batches_0000000001_to_0000000020.parquet
• etc.