نفرت_توهین آمیز

ds = tfds.load('huggingface:hate_offensive')

This dataset contains annotated tweets for automated hate-speech recognition

{
    "total_annotation_count": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "hate_speech_annotations": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "offensive_language_annotations": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "neither_annotations": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "num_classes": 3,
        "names": [
            "hate-speech",
            "offensive-language",
            "neither"
        ],
        "names_file": null,
        "id": null,
        "_type": "ClassLabel"
    },
    "tweet": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}