--- license: gpl pretty_name: tasksource/implicatures (mirror) source_datasets: - tasksource/implicatures tags: - laya - laya-finetune - mirror - domain-commonsense - use-conditional - genre-dialogue - label-choice - annot-unknown - lang-en --- # tasksource/implicatures(複製) ## データの内容 英語の対話で、応答に含まれる会話の含意(conversational implicature)を選ばせるデータ。1 行が 1 件で、`context` と `response` の組に対して、正しい含意を `correct_implicature`、誤った含意を `incorrect_implicature` に持つ。原典は George と Mamidi(2020)の注釈付き対話データで、誤った含意の選択肢は生成したディストラクタで補っている。 ## データの概要 | 項目 | 値 | |---|---| | 行数 | 1,000(1 構成・1 分割の合計) | | 列数 | 4 | | 容量 | 115.06 KB(ファイル 3 本) | | ファイルの形式 | csv 1 | | 公開 | public | ### 構成と分割 | 構成(config) | 分割(split)ごとの行数 | 列数 | |---|---|---:| | `default` | `train` 1,000 | 4 | ### 列(構成 `default`・分割 `train`) | 列 | 型 | 先頭の行の値 | |---|---|---| | `context` | `string` | `I keep putting off getting my passport applicati…` | | `response` | `string` | `Thank Goodness I did not drag my feet on that on…` | | `correct_implicature` | `string` | `I already sent my passport application` | | `incorrect_implicature` | `string` | `I did not yet send my passport application` | 行数と列は HF のデータセットビューアの集計(元のデータセットの現在の版で数えたもの)。全データセットの一覧は [HYUKI5/dataset-catalog](https://huggingface.co/datasets/HYUKI5/dataset-catalog) にある。 ## 複製について [`tasksource/implicatures`](https://huggingface.co/datasets/tasksource/implicatures) をリビジョン `f1ce91d0930acf52281380362961b75199afb2e5` の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(`convaiinnovations/laya`)の微調整に使う候補として、調査メモ(`Laya 学習用データセット調査`)に載せたデータを集めている。 ## 出典 - 元のデータセット: https://huggingface.co/datasets/tasksource/implicatures - 複製したリビジョン: `f1ce91d0930acf52281380362961b75199afb2e5`(2026-10-16 07:59:49+00:00) - 元のライセンス表記: `gpl` ## 論文 - https://arxiv.org/abs/2301.05948 ## アノテーションの方法 (元のカードに該当の見出しが無いので、ラベルの付け方に触れた段落を抜き出した) Implicature コーパス ```bib @article{george2020conversational, title={Conversational implicatures in English dialogue: Annotated dataset}, author={George, Elizabeth Jasmi and Mamidi, Radhika}, journal={Procedia Computer Science}, volume={171}, pages={2316--2323}, year={2020}, publisher={Elsevier} } ``` 生成したディストラクタで拡張した(tasksource 用): https://colab.research.google.com/drive/1ix0FgwzPAjQkIQA2E3ctlylvcmya7vGy?usp=sharing ## 微調整に使うときの注意 - 調査メモの台帳での区分: B(継承条件・gated・モデル由来の条件・上流の規約などが付く)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。 - Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。 - 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。 ## 調査メモの記述 | メモ | 節 | ライセンス(記載) | 確認 | 注記 | |---|---|---|---|---| | `jev_sources_tasksource_bekko.csv` | CSV(tasksource-jev-typed-decisions) | gpl / license_use=commercial | 確認済 | license_use=commercial / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r | ## 元のカード(日本語訳) 以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。 --- Implicature コーパス ```bib @article{george2020conversational, title={Conversational implicatures in English dialogue: Annotated dataset}, author={George, Elizabeth Jasmi and Mamidi, Radhika}, journal={Procedia Computer Science}, volume={171}, pages={2316--2323}, year={2020}, publisher={Elsevier} } ``` 生成したディストラクタで拡張した(tasksource 用): https://colab.research.google.com/drive/1ix0FgwzPAjQkIQA2E3ctlylvcmya7vGy?usp=sharing ```bib @article{sileo2023tasksource, title={tasksource: Structured Dataset Preprocessing Annotations for Frictionless Extreme Multi-Task Learning and Evaluation}, author={Sileo, Damien}, url= {https://arxiv.org/abs/2301.05948}, journal={arXiv preprint arXiv:2301.05948}, year={2023} } ```
原文(英語) Implicature corpus ```bib @article{george2020conversational, title={Conversational implicatures in English dialogue: Annotated dataset}, author={George, Elizabeth Jasmi and Mamidi, Radhika}, journal={Procedia Computer Science}, volume={171}, pages={2316--2323}, year={2020}, publisher={Elsevier} } ``` Augmented with generated distractors https://colab.research.google.com/drive/1ix0FgwzPAjQkIQA2E3ctlylvcmya7vGy?usp=sharing, for tasksource ```bib @article{sileo2023tasksource, title={tasksource: Structured Dataset Preprocessing Annotations for Frictionless Extreme Multi-Task Learning and Evaluation}, author={Sileo, Damien}, url= {https://arxiv.org/abs/2301.05948}, journal={arXiv preprint arXiv:2301.05948}, year={2023} } ```