bot_adversarial_dialogue
コレクションでコンテンツを整理
必要に応じて、コンテンツの保存と分類を行います。
ボット敵対対話データセット。
ボット敵対対話タスクからの攻撃性のラベルが付けられた対話データセット。対話は、人間にボットに敵対的に話しかけるよう依頼して収集されました。
詳細については論文をご覧ください。
@misc{xu2021recipes,
title={Recipes for Safety in Open-domain Chatbots},
author={Jing Xu and Da Ju and Margaret Li and Y-Lan Boureau and Jason Weston and Emily Dinan},
year={2021},
eprint={2010.07079},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
bot_adversarial_dialogue/dialogue_datasets (デフォルト設定)
スプリット | 例 |
---|
'test' | 2,598 |
'train' | 69,274 |
'valid' | 7,002 |
FeaturesDict({
'bot_persona': Sequence(Text(shape=(), dtype=string)),
'dialogue_id': float32,
'episode_done': bool,
'id': Text(shape=(), dtype=string),
'labels': ClassLabel(shape=(), dtype=int64, num_classes=2),
'round_id': float32,
'speaker_to_eval': Text(shape=(), dtype=string),
'text': Text(shape=(), dtype=string),
})
特徴 | クラス | 形 | Dタイプ | 説明 |
---|
| 特徴辞書 | | | |
ボットペルソナ | シーケンス(テキスト) | (なし、) | 弦 | ボットによって偽装されたペルソナ。 |
ダイアログID | テンソル | | float32 | |
エピソード_完了 | テンソル | | ブール | |
ID | 文章 | | 弦 | サンプルのID。 |
ラベル | クラスラベル | | int64 | |
ラウンドID | テンソル | | float32 | |
スピーカーから評価へ | 文章 | | 弦 | ラベルが付けられた発話の話者。 |
文章 | 文章 | | 弦 | 分類する発話。 |
bot_adversarial_dialogue/human_nonadv_safety_eval
FeaturesDict({
'episode_done': bool,
'id': Text(shape=(), dtype=string),
'labels': ClassLabel(shape=(), dtype=int64, num_classes=2),
'text': Text(shape=(), dtype=string),
})
特徴 | クラス | 形 | Dタイプ | 説明 |
---|
| 特徴辞書 | | | |
エピソード_完了 | テンソル | | ブール | |
ID | 文章 | | 弦 | サンプルのID。 |
ラベル | クラスラベル | | int64 | |
文章 | 文章 | | 弦 | 分類する発話。 |
特に記載のない限り、このページのコンテンツはクリエイティブ・コモンズの表示 4.0 ライセンスにより使用許諾されます。コードサンプルは Apache 2.0 ライセンスにより使用許諾されます。詳しくは、Google Developers サイトのポリシーをご覧ください。Java は Oracle および関連会社の登録商標です。
最終更新日 2023-09-09 UTC。
[null,null,["最終更新日 2023-09-09 UTC。"],[],[],null,["# bot_adversarial_dialogue\n\n\u003cbr /\u003e\n\n- **Description**:\n\nBot Adversarial Dialogue Dataset.\n=================================\n\nDialogue datasets labeled with offensiveness from Bot Adversarial Dialogue task.\nThe dialogues were collected by asking humans to adversarially talk to bots.\n\nMore details in the [paper](https://arxiv.org/abs/2010.07079).\n\n- **Homepage** :\n \u003chttps://github.com/facebookresearch/ParlAI/tree/main/parlai/tasks/bot_adversarial_dialogue\u003e\n\n- **Source code** :\n [`tfds.datasets.bot_adversarial_dialogue.Builder`](https://github.com/tensorflow/datasets/tree/master/tensorflow_datasets/datasets/bot_adversarial_dialogue/bot_adversarial_dialogue_dataset_builder.py)\n\n- **Versions**:\n\n - **`1.0.0`** (default): Initial release.\n- **Auto-cached**\n ([documentation](https://www.tensorflow.org/datasets/performances#auto-caching)):\n Yes\n\n- **Supervised keys** (See\n [`as_supervised` doc](https://www.tensorflow.org/datasets/api_docs/python/tfds/load#args)):\n `None`\n\n- **Figure**\n ([tfds.show_examples](https://www.tensorflow.org/datasets/api_docs/python/tfds/visualization/show_examples)):\n Not supported.\n\n- **Citation**:\n\n @misc{xu2021recipes,\n title={Recipes for Safety in Open-domain Chatbots},\n author={Jing Xu and Da Ju and Margaret Li and Y-Lan Boureau and Jason Weston and Emily Dinan},\n year={2021},\n eprint={2010.07079},\n archivePrefix={arXiv},\n primaryClass={cs.CL}\n }\n\nbot_adversarial_dialogue/dialogue_datasets (default config)\n-----------------------------------------------------------\n\n- **Config description**: The dialogue datasets, divided in train, validation\n and test splits.\n\n- **Download size** : `3.06 MiB`\n\n- **Dataset size** : `23.38 MiB`\n\n- **Splits**:\n\n| Split | Examples |\n|-----------|----------|\n| `'test'` | 2,598 |\n| `'train'` | 69,274 |\n| `'valid'` | 7,002 |\n\n- **Feature structure**:\n\n FeaturesDict({\n 'bot_persona': Sequence(Text(shape=(), dtype=string)),\n 'dialogue_id': float32,\n 'episode_done': bool,\n 'id': Text(shape=(), dtype=string),\n 'labels': ClassLabel(shape=(), dtype=int64, num_classes=2),\n 'round_id': float32,\n 'speaker_to_eval': Text(shape=(), dtype=string),\n 'text': Text(shape=(), dtype=string),\n })\n\n- **Feature documentation**:\n\n| Feature | Class | Shape | Dtype | Description |\n|-----------------|----------------|---------|---------|----------------------------------------|\n| | FeaturesDict | | | |\n| bot_persona | Sequence(Text) | (None,) | string | The persona impersonated by the bot. |\n| dialogue_id | Tensor | | float32 | |\n| episode_done | Tensor | | bool | |\n| id | Text | | string | The id of the sample. |\n| labels | ClassLabel | | int64 | |\n| round_id | Tensor | | float32 | |\n| speaker_to_eval | Text | | string | The speaker of the utterances labeled. |\n| text | Text | | string | The utterance to classify. |\n\n- **Examples** ([tfds.as_dataframe](https://www.tensorflow.org/datasets/api_docs/python/tfds/as_dataframe)):\n\nDisplay examples... \n\nbot_adversarial_dialogue/human_nonadv_safety_eval\n-------------------------------------------------\n\n- **Config description**: An human safety evaluation set evaluated by\n crowdsourced workers for offensiveness.\n\n- **Download size** : `10.57 KiB`\n\n- **Dataset size** : `34.55 KiB`\n\n- **Splits**:\n\n| Split | Examples |\n|----------|----------|\n| `'test'` | 180 |\n\n- **Feature structure**:\n\n FeaturesDict({\n 'episode_done': bool,\n 'id': Text(shape=(), dtype=string),\n 'labels': ClassLabel(shape=(), dtype=int64, num_classes=2),\n 'text': Text(shape=(), dtype=string),\n })\n\n- **Feature documentation**:\n\n| Feature | Class | Shape | Dtype | Description |\n|--------------|--------------|-------|--------|----------------------------|\n| | FeaturesDict | | | |\n| episode_done | Tensor | | bool | |\n| id | Text | | string | The id of the sample. |\n| labels | ClassLabel | | int64 | |\n| text | Text | | string | The utterance to classify. |\n\n- **Examples** ([tfds.as_dataframe](https://www.tensorflow.org/datasets/api_docs/python/tfds/as_dataframe)):\n\nDisplay examples..."]]