codec/scripts/compute_wer.py

import argparse
import json
import string

from jiwer import wer


def normalize_text(text: str) -> str:
    """
    Lowercase and remove punctuation from a string.

    Args:
        text (str): Input string

    Returns:
        str: Normalized string
    """
    # Lowercase
    text = text.lower()
    # Remove punctuation
    text = text.translate(str.maketrans("", "", string.punctuation))
    return text


def load_transcripts(jsonl_path):
    originals = []
    reconstructions = []
    with open(jsonl_path, "r", encoding="utf-8") as f:
        for line in f:
            data = json.loads(line)
            originals.append(data["original_text"])
            reconstructions.append(data["reconstructed_text"])
    return originals, reconstructions


def main(args):
    originals, reconstructions = map(normalize_text, load_transcripts(args.jsonl))
    score = wer(originals, reconstructions)
    print(f"WER: {score:.3%}")


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument(
        "--jsonl", type=str, required=True, help="Path to the transcript JSONL file"
    )
    args = parser.parse_args()
    main(args)