Book corpus datasets used to train large language models

{
    "object_type": "df_transaction_reference",
    "event_type": "DATASET_LICENSING",
    "buyer": "Microsoft",
    "seller": "HarperCollins",
    "asset_description": "Book corpus datasets used to train large language models",
    "reported_price": "$1,667 per book title annually",
    "confidence": "Medium-High",
    "region": "Global",
    "_attribution": {
        "source": "DatFlash",
        "publisher": "DataUniversa",
        "retrieved": "2024",
        "note": "Observed transaction reference. No normalization or appraisal applied."
    }
}