vocalization.py

"""
Module for vocalization of Arabic text
"""

# Using mishkal
# See: https://github.com/linuxscout/mishkal?tab=readme-ov-file#example

# import json

from contextlib import suppress

import mishkal.tashkeel

# import requests
import data
from arab_tools import join, separate

vocalizer = mishkal.tashkeel.TashkeelClass("")


# def vocalize(text: str) -> str:
#     o = vocalizer.tashkeel(text)
#     o = "".join(c if c.isprintable() else " " for c in o).strip()
#     return o


# This doesn't work
# def vocalize(text: str) -> str:
#     task = "diacritization"
#     data = {"text": text, "task": task, "API_KEY": "lpcsTkDIDf"}
#     response = requests.post(
#         "https://farasa.qcri.org/diacritization/analyze/",
#         data=data,
#     )
#     response.raise_for_status()
#     return json.loads(response.text)["text"]


def _find_haraka(word: str):
    for c in word:
        if c in data.harakat_wo_shaddah:
            return c
    return ""


def vocalize(text: str) -> str:
    from gradio_client import Client

    client = Client("https://testingdoang-shakkala-arabic-tashkeel.hf.space/")
    new_text = client.predict(text, api_name="/predict")
    rasm, old = separate(text)
    _, new = separate(new_text)
    # assert rasm == _
    s = data.shaddah
    joint_harakat = [
        (s if s in o or s in n else "") + (_find_haraka(o) or _find_haraka(n))
        for o, n in zip(old, new)
    ]
    return join(rasm, joint_harakat)


# fix asmai from here

with suppress(ImportError):
    import asmai.semdictionary  # noqa

    asmai.semdictionary.SemanticDictionary.lookup_rule = (
        lambda self, primate_word, second_word: data.sem_relations.get(
            (primate_word, second_word), False
        )
    )