Dataset Card for luotuo-QA-B

Dataset Summary

Anki_Card是一种用于记忆和学习的电子卡片系统。我们建立了一个类似于这种形式的问答数据集，旨在推动中英文语境下问答模型的研究和发展。

我们的数据集是在3个开源数据集之上生成构建的，这3个数据集分别是：

·Chinese Scientific Literature Dataset

·CNN-DailyMail News Text Summarization

·arXiv Dataset

您可以直接搜索这些原始数据集的名称或是从以下链接访问它们

·https://github.com/ydli-ai/CSL

·https://www.kaggle.com/datasets/gowrishankarp/newspaper-text-summarization-cnn-dailymail

·https://www.kaggle.com/datasets/Cornell-University/arxiv

我们在这些数据集的基础上针对每一个摘要或新闻生成了5个“问题-答案”对。数据分布如下：

---从Chinese Scientific Literature Dataset(CSL)数据集中生成了25836条中文数据，共129180个问答对。

---从CNN-DailyMail News Text Summarization数据集中生成了2026条数据，共10130个问答对。

---从arXiv Dataset数据集中生成了3602条英文数据，共18010个问答对。

此外，由于此数据集是我们Luotuo-QA项目的一部分，我们将它叫做luotuo-QA-B。

您可以在这里查看Luotuo-QA项目：https://github.com/LC1332/Luotuo-QA

此数据集适用于训练和评估中文对话式问答模型。有益于推动中文自然语言处理领域的发展，同时也为研究人员和开发者提供了一个基准，用于比较不同模型的性能和探索新的方法。

我们希望这一工作能够促进全球范围内中文语境对话式问答任务的研究和进一步的创新。

Anki_Card is an electronic flashcard system used for memory and learning. We have created a question-and-answer dataset in a similar format to facilitate research and development of question-answering models in both Chinese and English contexts.

Our dataset is constructed based on three open-source datasets:

·Chinese Scientific Literature Dataset

·CNN-DailyMail News Text Summarization

·arXiv Dataset

You can directly search for the names of these original datasets or access them from the following links:

·Chinese Scientific Literature Dataset (CSL): https://github.com/ydli-ai/CSL

·CNN-DailyMail News Text Summarization: https://www.kaggle.com/datasets/gowrishankarp/newspaper-text-summarization-cnn-dailymail

·arXiv Dataset: https://www.kaggle.com/datasets/Cornell-University/arxiv

Based on these datasets, we have generated five "question-answer" pairs for each summary or news article. The data distribution is as follows:

---From the Chinese Scientific Literature Dataset (CSL), we generated 25,836 Chinese data points, resulting in a total of 129,180 question-answer pairs.

---From the CNN-DailyMail News Text Summarization dataset, we generated 2,026 data points, resulting in a total of 10,130 question-answer pairs.

---From the arXiv Dataset, we generated 3,602 English data points, resulting in a total of 18,010 question-answer pairs.

Furthermore, as this dataset is part of our Luotuo-QA project, we refer to it as luotuo-QA-B.

You can find the Luotuo-QA project here: https://github.com/LC1332/Luotuo-QA

This dataset is suitable for training and evaluating Chinese conversational question-answering models. It contributes to the development of Chinese natural language processing and provides researchers and developers with a benchmark for comparing the performance of different models and exploring new approaches.

We hope that this work will promote research and further innovation in Chinese conversational question-answering tasks on a global scale.

Languages

CHINESE, ENGLISH

Data Instances

中文数据样例：

{
  "story": "中国股市发展中特有的股权分置结构决定了研究股市收益率问题的复杂性.本文提出用全收益率的标准来衡量中国股市的整体收益率,认为在股权分置及其逐步解决的过程中,研究股市全收益率具有重要的意义,也是讨论股市其它问题的理论基础.随着股权分置改革渐进式地推进,中国股市各类股权所有者的收益分布会发生显著的结构性变化.从长期看,股权分置改革能使投资股东和原始股东的收益函数趋于一致,有助于实现整体收益的最大化.",
  "questions": [
    "为什么研究股市收益率问题复杂？",
    "用什么标准来衡量中国股市的整体收益率？",
    "股权分置改革对股东收益分布会有什么影响？",
    "股权分置改革的推进方式是什么？",
    "为什么研究股市全收益率具有重要意义？"
  ],
  "answers": [
    "因为中国股市发展中特有的股权分置结构决定了研究股市收益率问题的复杂性。",
    "用全收益率的标准来衡量中国股市的整体收益率。",
    "股权分置改革会使投资股东和原始股东的收益函数趋于一致，有助于实现整体收益的最大化。",
    "股权分置改革是渐进式地推进的。",
    "因为研究股市全收益率是讨论股市其它问题的理论基础，也在股权分置及其逐步解决的过程中具有重要的意义。"
  ],
  "language": "Chinese"
}

英文数据样例：

{
  "story": "'(CNN) -- A 14-year-old was arrested late Tuesday after shining a powerful laser light into the eyes of a pilot who was approaching Los Angeles International Airport, the Federal Aviation Administration said. The arrest puts a spotlight on what the FAA calls a dangerous problem in recent years. In Tuesday's case, the pilot was about 2,000 feet in the air and nobody was hurt in the incident, said Ian Gregor, an FAA spokesman. \"It's potentially very dangerous to shine a laser at an aircraft because a laser can distract a pilot and there have been cases where pilots have suffered temporary vision problems as a result of being struck by a laser beam,\" Gregor said. \" We've had reports of pilots having to turn over control of the aircraft to a co-pilot or had to abort landing.\" Gregor said Los Angeles International Airport has had many instances of laser attacks. \"Pilots reported 102 laser incidents around LAX in 2010. Most of any airport in the country,\" Gregor said.'",
  "questions": [
    "What happened to the 14-year-old?",
    "Why is shining a laser at an aircraft dangerous?",
    "What have pilots had to do in some cases of laser attacks?",
    "How many laser incidents were reported around LAX in 2010?",
    "What is the FAA's concern about laser attacks?"
  ],
  "answers": [
    "The 14-year-old was arrested for shining a powerful laser light into the eyes of a pilot.",
    "Shining a laser at an aircraft is dangerous because it can distract a pilot and cause temporary vision problems.",
    "In some cases of laser attacks, pilots have had to turn over control of the aircraft to a co-pilot or had to abort landing.",
    "102 laser incidents were reported around LAX in 2010, the most of any airport in the country.",
    "The FAA is concerned about laser attacks because they pose a dangerous problem for pilots and can cause temporary vision problems."
  ],
  "language": "English"
}

Licensing Information

我们的协议与三个原始数据集的协议保持一致，请阅读以下内容。

·CSL数据集的协议是Apache License 2.0，除非遵守许可证，否则您不得使用此文件

·CNN-DailyMail News Text Summarization数据集的协议是 CC0: Public Domain

·arXiv数据集的协议是 CC0: Public Domain

Our agreements are consistent with the agreements of three original datasets. Please read the following information.

· The protocol for the CSL dataset is Apache License 2.0. You are not allowed to use this file unless you comply with the license.

· The protocol for the CNN-DailyMail News Text Summarization dataset is CC0: Public Domain.

· The protocol for the arXiv dataset is CC0: Public Domain.

Citation Information

如果您在项目中使用了我们的模型、代码或者数据，请引用我们。

Please cite us if you use the data or code in this repo.

@misc{alpaca,
  author={Jianshen Liao, Ao Sun, Qinyu Luo, Hongsen Huang, Cheng Li},
  title = {Luotuo-QA: Better Conversational Question Answering Model with Answer Completion},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/LC1332/Luotuo-QA}},
}

Datasets:

Logic123456789
/

Luotuo-QA-B