---
dataset_info:
features:
- name: conversations
list:
- name: from
dtype: string
- name: value
dtype: string
- name: id
dtype: string
splits:
- name: train
num_bytes: 4550239762
num_examples: 3606402
download_size: 2796036198
dataset_size: 4550239762
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
task_categories:
- text2text-generation
language:
- zh
size_categories:
- 1M<n<10M
---
# Dataset Card for "train_3.5m-zhtw"
## 内容
包含約 350 萬條由 [BELLE](https://github.com/LianjiaTech/BELLE) 計畫產生的中文指令資料。
## 範例
```
{
"id":"66623028",
"conversations":[
{
"from":"human",
"value":"歸納介紹三種不同類型的電影專題"
},
{
"from":"assistant",
"value":"1. 動作片:動作片通常強調動作和戰鬥場面,故事情節緊湊,充滿了緊張和興奮的氛圍。主角一般是具有超凡實力或特殊技能的英雄人物,往往會與反派角色進行一系列激烈的格鬥,追逐和爆破等酷炫場面。例如《玩命關頭》和《復仇者聯盟》等。\n2. 愛情片:愛情片是以愛情為主題的電影,著重呈現角色之間的感情糾葛,展現了愛情的美麗與悲壯。情節與角色之間的發展令人深思,增加了觀眾的感性體驗。這種電影的故事情節主要以人類與人類之間的情感變化為主題。例如《泰坦尼克號》和《讓子彈飛》等。\n3. 科幻片:科幻片是以未來科技和未知世界為主題的電影,通常包括時間旅行、外星人、虛擬實境、未來社會等各種奇幻的元素。這種電影描繪了一種比現實更加宏偉、更加神秘和驚奇的世界。例如《星際效應》和《發條橘子》等。"
}
]
}
{
"id":"18205437",
"conversations":[
{
"from":"human",
"value":"根據給定的食材,編寫一道美食食譜。\n雞腿肉、蘑菇、洋蔥"
},
{
"from":"assistant",
"value":"雞肉炒蘑菇洋蔥\n食材:\n- 雞腿肉400克,去骨去皮切成小塊\n- 蘑菇200克,切片\n- 洋蔥1個,切絲\n- 生薑適量,切絲\n- 醬油2匙\n- 料酒1匙\n- 鹽適量\n- 白胡椒粉適量\n- 植物油適量\n做法:\n1. 在一個碗裡放入切好的雞腿肉塊,加入醬油、料酒和適量的白胡椒粉,攪拌均勻,醃製10分鐘。\n2. 熱鍋涼油,加入薑絲翻炒,放入醃好的雞肉塊翻炒至變色。\n3 . 加入洋蔥絲和蘑菇,翻炒2-3分鐘至蘑菇變軟。\n4. 加入適量的鹽和少量的水,蓋上鍋蓋燜煮5分鐘直至雞肉熟透。\n5. 最後翻炒幾下,使湯汁變濃稠即可。\n溫馨提示:\n1. 在醃製雞肉時,如果想讓肉質更嫩滑,可以加入少量的澱粉和蛋清。\n2. 可以根據個人口味加入其他調味料,如蒜末、辣椒等。\n3. 洋蔥和蘑菇可以用其他蔬菜代替,如青椒、胡蘿蔔等。"
},
{
"from":"human",
"value":"太好了,我會試著照著你教的做一次,但是我最近買了一部新相機,想讓你推荐一些好的拍照地點。"
},
{
"from":"assistant",
"value":"當地的風景區、公園、景點等等都是比較適合拍照的地點。如果你想要一些特別的地方,可以考慮一些建築群,街道巷子、咖啡館、書店等具有文藝氣息的地方。"
},
{
"from":"human",
"value":"嗯,這些地方都很不錯,但是我還想找一些可以體驗當地文化和風俗的地點,你有什麼好的建議嗎?"
},
{
"from":"assistant",
"value":"如果你想要更深入地了解當地文化和風俗,可以考慮參加一些遊學活動,例如參觀當地的博物館、參加民俗節慶、嘗試當地美食等等。此外,也可以去一些小城鎮,感受那裡的古樸風情。"
}
]
}
```
### 欄位:
```
id: 資料id
conversations: 數據內容,以對話形式給出,包括多輪和單輪對話的數據
```
### 補充:
針對此資料集,新增了指令類別字段,共包括13個類別,詳情見
[train_3.5M_CN_With_Category](https://huggingface.co/datasets/BELLE-2/train_3.5M_CN_With_Category)
## 使用限制
僅允許將此資料集及使用此資料集產生的衍生物用於研究目的,不得用於商業,以及其他會對社會帶來危害的用途。
本資料集不代表任何一方的立場、利益或想法,無關任何團體的任何類型的主張。因使用本資料集所帶來的任何損害、糾紛,本專案不承擔任何責任。
## Citation
```bibtex
@misc{BELLE,
author = {Yunjie Ji, Yong Deng, Yan Gong, Yiping Peng, Qiang Niu, Baochang Ma and Xiangang Li},
title = {BELLE: Be Everyone's Large Language model Engine },
year = {2023},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/LianjiaTech/BELLE}},
}
@article{belle2023exploring,
title={Exploring the Impact of Instruction Data Scaling on Large Language Models: An Empirical Study on Real-World Use Cases},
author={Yunjie Ji, Yong Deng, Yan Gong, Yiping Peng, Qiang Niu, Lei Zhang, Baochang Ma, Xiangang Li},
journal={arXiv preprint arXiv:2303.14742},
year={2023}
}
```
---
dataset_info:
features:
- name: conversations
list:
- name: from
dtype: string
- name: value
dtype: string
- name: id
dtype: string
splits:
- name: train
num_bytes: 4550239762
num_examples: 3606402
download_size: 2796036198
dataset_size: 4550239762
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
task_categories:
- text2text-generation
language:
- zh
size_categories:
- 1M<n<10M
---
# Dataset Card for "train_3.5m-zhtw"
## Content
Contains approximately 3.5 million Chinese instruction datasets generated by the [BELLE](https://github.com/LianjiaTech/BELLE) project.
## Examples
{
"id":"66623028",
"conversations":[
{
"from":"human",
"value":"歸納介紹三種不同類型的電影專題"
},
{
"from":"assistant",
"value":"1. 動作片:動作片通常強調動作和戰鬥場面,故事情節緊湊,充滿了緊張和興奮的氛圍。主角一般是具有超凡實力或特殊技能的英雄人物,往往會與反派角色進行一系列激烈的格鬥,追逐和爆破等酷炫場面。例如《玩命關頭》和《復仇者聯盟》等。
2. 愛情片:愛情片是以愛情為主題的電影,著重呈現角色之間的感情糾葛,展現了愛情的美麗與悲壯。情節與角色之間的發展令人深思,增加了觀眾的感性體驗。這種電影的故事情節主要以人類與人類之間的情感變化為主題。例如《泰坦尼克號》和《讓子彈飛》等。
3. 科幻片:科幻片是以未來科技和未知世界為主題的電影,通常包括時間旅行、外星人、虛擬實境、未來社會等各種奇幻的元素。這種電影描繪了一種比現實更加宏偉、更加神秘和驚奇的世界。例如《星際效應》和《發條橘子》等。"
}
]
}
{
"id":"18205437",
"conversations":[
{
"from":"human",
"value":"根據給定的食材,編寫一道美食食譜。
雞腿肉、蘑菇、洋蔥"
},
{
"from":"assistant",
"value":"雞肉炒蘑菇洋蔥
食材:
- 雞腿肉400克,去骨去皮切成小塊
- 蘑菇200克,切片
- 洋蔥1個,切絲
- 生薑適量,切絲
- 醬油2匙
- 料酒1匙
- 鹽適量
- 白胡椒粉適量
- 植物油適量
做法:
1. 在一個碗裡放入切好的雞腿肉塊,加入醬油、料酒和適量的白胡椒粉,攪拌均勻,醃製10分鐘。
2. 熱鍋涼油,加入薑絲翻炒,放入醃好的雞肉塊翻炒至變色。
3 . 加入洋蔥絲和蘑菇,翻炒2-3分鐘至蘑菇變軟。
4. 加入適量的鹽和少量的水,蓋上鍋蓋燜煮5分鐘直至雞肉熟透。
5. 最後翻炒幾下,使湯汁變濃稠即可。
溫馨提示:
1. 在醃製雞肉時,如果想讓肉質更嫩滑,可以加入少量的澱粉和蛋清。
2. 可以根據個人口味加入其他調味料,如蒜末、辣椒等。
3. 洋蔥和蘑菇可以用其他蔬菜代替,如青椒、胡蘿蔔等。"
},
{
"from":"human",
"value":"太好了,我會試著照著你教的做一次,但是我最近買了一部新相機,想讓你推荐一些好的拍照地點。"
},
{
"from":"assistant",
"value":"當地的風景區、公園、景點等等都是比較適合拍照的地點。如果你想要一些特別的地方,可以考慮一些建築群,街道巷子、咖啡館、書店等具有文藝氣息的地方。"
},
{
"from":"human",
"value":"嗯,這些地方都很不錯,但是我還想找一些可以體驗當地文化和風俗的地點,你有什麼好的建議嗎?"
},
{
"from":"assistant",
"value":"如果你想要更深入地了解當地文化和風俗,可以考慮參加一些遊學活動,例如參觀當地的博物館、參加民俗節慶、嘗試當地美食等等。此外,也可以去一些小城鎮,感受那裡的古樸風情。"
}
]
}
## Fields
id: Unique data identifier
conversations: Data content presented in conversational format, including both single-turn and multi-turn dialogue datasets
## Additional Notes
An instruction category field has been added to this dataset, totaling 13 categories. For details, please refer to [train_3.5M_CN_With_Category](https://huggingface.co/datasets/BELLE-2/train_3.5M_CN_With_Category)
## Usage Restrictions
This dataset and derivatives generated using it are only permitted for research purposes. Commercial use and other uses that may cause harm to society are prohibited. This dataset does not represent the positions, interests, or opinions of any party, and is unrelated to any type of claim of any organization. The project shall not be liable for any damages or disputes arising from the use of this dataset.
## Citation
bibtex
@misc{BELLE,
author = {Yunjie Ji, Yong Deng, Yan Gong, Yiping Peng, Qiang Niu, Baochang Ma and Xiangang Li},
title = {BELLE: Be Everyone's Large Language model Engine },
year = {2023},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {url{https://github.com/LianjiaTech/BELLE}},
}
@article{belle2023exploring,
title={Exploring the Impact of Instruction Data Scaling on Large Language Models: An Empirical Study on Real-World Use Cases},
author={Yunjie Ji, Yong Deng, Yan Gong, Yiping Peng, Qiang Niu, Lei Zhang, Baochang Ma, Xiangang Li},
journal={arXiv preprint arXiv:2303.14742},
year={2023}
}