e2e_cleaned

  • Mô tả :

Bản phát hành cập nhật dữ liệu Thử thách E2E NLG với các MR đã được làm sạch. Dữ liệu E2E chứa biểu diễn ý nghĩa dựa trên hành động đối thoại (MR) trong miền nhà hàng và tối đa 5 tham chiếu bằng ngôn ngữ tự nhiên, đây là điều mà người ta cần dự đoán.

Tách ra ví dụ
'test' 4,693
'train' 33,525
'validation' 4.299
  • Cấu trúc tính năng :
FeaturesDict({
    'input_text': FeaturesDict({
        'table': Sequence({
            'column_header': string,
            'content': string,
            'row_number': int16,
        }),
    }),
    'target_text': string,
})
  • Tài liệu tính năng :
Tính năng Lớp Hình dạng Dtype Sự mô tả
Tính năngDict
nhập ký tự Tính năngDict
input_text/bảng Sự phối hợp
input_text/table/column_header tenxơ sợi dây
input_text/bảng/nội dung tenxơ sợi dây
input_text/table/row_number tenxơ int16
văn bản đích tenxơ sợi dây
  • trích dẫn :
@inproceedings{dusek-etal-2019-semantic,
    title = "Semantic Noise Matters for Neural Natural Language Generation",
    author = "Du{\v{s} }ek, Ond{\v{r} }ej  and
      Howcroft, David M.  and
      Rieser, Verena",
    booktitle = "Proceedings of the 12th International Conference on Natural Language Generation",
    month = oct # "{--}" # nov,
    year = "2019",
    address = "Tokyo, Japan",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/W19-8652",
    doi = "10.18653/v1/W19-8652",
    pages = "421--426",
    abstract = "Neural natural language generation (NNLG) systems are known for their pathological outputs, i.e. generating text which is unrelated to the input specification. In this paper, we show the impact of semantic noise on state-of-the-art NNLG models which implement different semantic control mechanisms. We find that cleaned data can improve semantic correctness by up to 97{\%}, while maintaining fluency. We also find that the most common error is omitting information, rather than hallucination.",
}