טשקלע

ds = tfds.load('huggingface:tashkeela/plain_text')

Arabic vocalized texts.
it contains 75 million of fully vocalized words mainly97 books from classical and modern Arabic language.

{
    "text": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "book": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}