dokument2dial

Referencje:

domena_dialogu

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:doc2dial/dialogue_domain')
  • Opis :
Doc2dial is dataset of goal-oriented dialogues that are grounded in the associated documents. It includes over 4500 annotated conversations with an average of 14 turns that are grounded in over 450 documents from four domains. Compared to the prior document-grounded dialogue datasets this dataset covers a variety of dialogue scenes in information-seeking conversations.
  • Licencja : Brak znanej licencji
  • Wersja : 1.0.1
  • Podziały :
Podział Przykłady
'train' 3474
'validation' 661
  • Cechy :
{
    "dial_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "doc_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "domain": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "turns": [
        {
            "turn_id": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "role": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "da": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "references": [
                {
                    "sp_id": {
                        "dtype": "string",
                        "id": null,
                        "_type": "Value"
                    },
                    "label": {
                        "dtype": "string",
                        "id": null,
                        "_type": "Value"
                    }
                }
            ],
            "utterance": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        }
    ]
}

domena_dokumentu

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:doc2dial/document_domain')
  • Opis :
Doc2dial is dataset of goal-oriented dialogues that are grounded in the associated documents. It includes over 4500 annotated conversations with an average of 14 turns that are grounded in over 450 documents from four domains. Compared to the prior document-grounded dialogue datasets this dataset covers a variety of dialogue scenes in information-seeking conversations.
  • Licencja : Brak znanej licencji
  • Wersja : 1.0.1
  • Podziały :
Podział Przykłady
'train' 3416
  • Cechy :
{
    "domain": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "doc_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "doc_text": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "spans": [
        {
            "id_sp": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "tag": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "start_sp": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "end_sp": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text_sp": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "title": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "parent_titles": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "id_sec": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "start_sec": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text_sec": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "end_sec": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            }
        }
    ],
    "doc_html_ts": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "doc_html_raw": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

doc2dial_rc

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:doc2dial/doc2dial_rc')
  • Opis :
Doc2dial is dataset of goal-oriented dialogues that are grounded in the associated documents. It includes over 4500 annotated conversations with an average of 14 turns that are grounded in over 450 documents from four domains. Compared to the prior document-grounded dialogue datasets this dataset covers a variety of dialogue scenes in information-seeking conversations.
  • Licencja : Brak znanej licencji
  • Wersja : 1.0.1
  • Podziały :
Podział Przykłady
'train' 20431
'validation' 3972
  • Cechy :
{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "domain": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}