step7_fallback_phonemes_and_madd.py

import json
import re
from arabic_phonemizer import ArabicPhonemizer

# --- Helpers ---
# Very simple Madd detection from script (MVP-level):
# We mark likely long vowels caused by: ا, و, ي, ى, and madd sign "ٓ"
MADD_CHARS = set(["ا", "و", "ي", "ى", "ٓ"])

ARABIC_DIACRITICS = re.compile(r"[\u064B-\u0652\u0670\u0653\u0654\u0655]")  # tanwin, harakat, etc.

def strip_diacritics(s: str) -> str:
    return re.sub(ARABIC_DIACRITICS, "", s)

def detect_madd_positions(word: str):
    """
    Returns a list of indices in the *diacritics-stripped* word where Madd-ish characters appear.
    MVP heuristic; later replace with Quranic-Phonemizer (Tajweed-aware).
    """
    base = strip_diacritics(word)
    return [i for i, ch in enumerate(base) if ch in MADD_CHARS]

def main():
    # Instantiate phonemizer once
    ph = ArabicPhonemizer()

    path_in = "data/fatiha_canonical.json"
    with open(path_in, "r", encoding="utf-8") as f:
        data = json.load(f)

    for ay in data["ayahs"]:
        ay_word_info = []
        for w in ay["words"]:
            base = strip_diacritics(w)

            # ArabicPhonemizer API: use .phonemize(text)
            # If your version differs, we’ll adapt after you run it.
            phonemes = ph.phonemize(w)

            ay_word_info.append({
                "word": w,
                "base": base,
                "phonemes_fallback": phonemes,
                "madd_positions_base_index": detect_madd_positions(w)
            })
        ay["word_info"] = ay_word_info

    path_out = "data/fatiha_canonical_fallback.json"
    with open(path_out, "w", encoding="utf-8") as f:
        json.dump(data, f, ensure_ascii=False, indent=2)

    print("OK ✅ wrote", path_out)
    print("Sample ayah 1 word_info:")
    for item in data["ayahs"][0]["word_info"]:
        print(" -", item["word"], "| base:", item["base"], "| madd idx:", item["madd_positions_base_index"], "| ph:", item["phonemes_fallback"])

if __name__ == "__main__":
    main()