load_dataset.py

#!/usr/bin/env python3
"""
Example script to load and use the Common Voice Japanese test set.
"""

import json
from pathlib import Path
from collections import Counter

def load_cv_test_set(dataset_dir="."):
    """
    Load Common Voice Japanese test set from all splits.

    Args:
        dataset_dir: Path to the dataset directory

    Returns:
        List of dicts with audio paths and metadata
    """
    dataset_dir = Path(dataset_dir)
    samples = []

    # Load from all splits (ja_00, ja_01, ja_02, ja_03)
    for split_dir in sorted(dataset_dir.glob("ja_[0-9][0-9]")):
        metadata_file = split_dir / "test.jsonl"

        if not metadata_file.exists():
            continue

        with open(metadata_file, 'r', encoding='utf-8') as f:
            for line in f:
                entry = json.loads(line)
                entry['audio_path'] = str(split_dir / entry['path'])
                entry['split'] = split_dir.name
                samples.append(entry)

    return samples


def get_dataset_stats(samples):
    """Get statistics about the dataset."""
    stats = {
        'total_samples': len(samples),
        'gender_distribution': Counter(s.get('gender', '') for s in samples),
        'age_distribution': Counter(s.get('age', '') for s in samples),
        'split_distribution': Counter(s.get('split', '') for s in samples),
        'total_votes': sum(s.get('up_votes', 0) + s.get('down_votes', 0) for s in samples),
        'avg_up_votes': sum(s.get('up_votes', 0) for s in samples) / len(samples) if samples else 0,
    }
    return stats


if __name__ == "__main__":
    # Load test set
    print("Loading Common Voice Japanese test set...\n")

    samples = load_cv_test_set()
    stats = get_dataset_stats(samples)

    print(f"=== Test Set Statistics ===")
    print(f"Total samples: {stats['total_samples']}")
    print(f"Average up votes: {stats['avg_up_votes']:.2f}")

    print("\nSplit distribution:")
    for split, count in sorted(stats['split_distribution'].items()):
        print(f"  {split}: {count} files")

    print("\nGender distribution:")
    for gender, count in stats['gender_distribution'].most_common():
        if gender:
            print(f"  {gender}: {count} ({count/stats['total_samples']*100:.1f}%)")

    print("\nAge distribution:")
    for age, count in stats['age_distribution'].most_common():
        if age:
            print(f"  {age}: {count} ({count/stats['total_samples']*100:.1f}%)")

    # Show sample
    if samples:
        print(f"\nSample entry:")
        sample = samples[0]
        print(f"  Split: {sample['split']}")
        print(f"  File: {sample['file_name']}")
        print(f"  Path: {sample['path']}")
        print(f"  Text: {sample['text'][:50]}...")
        print(f"  Gender: {sample.get('gender', 'N/A')}")
        print(f"  Age: {sample.get('age', 'N/A')}")