sedthh/gutenberg_english
收藏资源简介:
--- dataset_info: features: - name: TEXT dtype: string - name: SOURCE dtype: string - name: METADATA dtype: string splits: - name: train num_bytes: 18104255935 num_examples: 48284 download_size: 10748877194 dataset_size: 18104255935 license: mit task_categories: - text-generation language: - en tags: - project gutenberg - e-book - gutenberg.org pretty_name: Project Gutenberg eBooks in English size_categories: - 10K<n<100K --- # Dataset Card for Project Gutenber - English Language eBooks A collection of non-english language eBooks (48284 rows, 80%+ of all english language books available on the site) from the Project Gutenberg site with metadata removed. Originally colected for https://github.com/LAION-AI/Open-Assistant (follows the OpenAssistant training format) The METADATA column contains catalogue meta information on each book as a serialized JSON: | key | original column | |----|----| | language | - | | text_id | Text# unique book identifier on Prject Gutenberg as *int* | | title | Title of the book as *string* | | issued | Issued date as *string* | | authors | Authors as *string*, comma separated sometimes with dates | | subjects | Subjects as *string*, various formats | | locc | LoCC code as *string* | | bookshelves | Bookshelves as *string*, optional | ## Source data **How was the data generated?** - A crawler (see Open-Assistant repository) downloaded the raw HTML code for each eBook based on **Text#** id in the Gutenberg catalogue (if available) - The metadata and the body of text are not clearly separated so an additional parser attempts to split them, then remove transcriber's notes and e-book related information from the body of text (text clearly marked as copyrighted or malformed was skipped and not collected) - The body of cleaned TEXT as well as the catalogue METADATA is then saved as a parquet file, with all columns being strings **Copyright notice:** - Some of the books are copyrighted! The crawler ignored all books with an english copyright header by utilizing a regex expression, but make sure to check out the metadata for each book manually to ensure they are okay to use in your country! More information on copyright: https://www.gutenberg.org/help/copyright.html and https://www.gutenberg.org/policy/permission.html - Project Gutenberg has the following requests when using books without metadata: _Books obtianed from the Project Gutenberg site should have the following legal note next to them: "This eBook is for the use of anyone anywhere in the United States and most other parts of the world at no cost and with almost" no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org. If you are not located in the United States, you will have to check the laws of the country where you are located before using this eBook."_
数据集信息: 特征字段: - 字段名:TEXT,数据类型:字符串 - 字段名:SOURCE,数据类型:字符串 - 字段名:METADATA,数据类型:字符串 数据集划分: - 划分名称:训练集(train),占用字节数:18104255935,样本数量:48284 下载大小:10748877194,数据集总大小:18104255935 开源协议:MIT许可证(mit) 任务类别: - 文本生成(text-generation) 语言: - 英语(en) 标签: - 古腾堡计划(Project Gutenberg) - 电子书(e-book) - 古腾堡官网(gutenberg.org) 美观名称:英语语言古腾堡计划电子书(Project Gutenberg eBooks in English) 规模类别: - 样本量介于10000至100000之间(10K<n<100K) # 古腾堡计划英语语言电子书数据集卡片 本数据集源自古腾堡计划(Project Gutenberg)官网,为移除元数据的英语电子书集合(共48284条样本,占该网站可获取英语书籍总量的80%以上)。 本数据集最初为https://github.com/LAION-AI/Open-Assistant 项目采集,遵循OpenAssistant训练格式。 METADATA字段以序列化JSON格式存储每本图书的编目元信息,具体键值说明如下表所示: | 元数据键 | 说明 | | ---- | ---- | | language | 语言 | | text_id | 古腾堡计划图书唯一标识符Text#,类型为整数(int) | | title | 图书标题,类型为字符串(string) | | issued | 图书发布日期,类型为字符串(string) | | authors | 作者信息,类型为字符串,多个作者以逗号分隔,部分条目附带作者生卒日期 | | subjects | 图书主题,类型为字符串,格式多样 | | locc | LoCC分类代码(Library of Congress Classification Code),类型为字符串(string) | | bookshelves | 书架分类信息,类型为字符串,为可选字段 | ## 数据源 ### 数据采集流程 - 爬虫(详见Open-Assistant仓库)基于古腾堡编目中的**Text#**编号(若存在)下载每本电子书的原始HTML代码 - 由于元数据与正文未明确分离,需通过额外解析器拆分二者,随后从正文中移除转录者注释及电子书相关信息(带有明确版权声明或格式错误的文本将被跳过,不予收录) - 清理后的正文TEXT字段与编目METADATA字段将保存为Parquet格式文件(Parquet),所有字段均为字符串类型 ## 版权声明 - 部分图书受版权保护!爬虫通过正则表达式过滤了带有英语版权声明的图书,但请务必手动检查每本图书的元数据,确保其在您所在国家可合法使用。更多版权相关信息请参考:https://www.gutenberg.org/help/copyright.html 与 https://www.gutenberg.org/policy/permission.html - 古腾堡计划对无元数据的图书使用提出如下要求:**"从古腾堡计划官网获取的电子书需附带以下法律声明:'本电子书可供美国境内及全球绝大多数地区的用户免费使用,几乎无任何使用限制。您可复制、分发或再使用本电子书,需遵循随本电子书附带的古腾堡计划许可协议,或访问www.gutenberg.org在线查看。若您不在美国境内,使用本电子书前请务必核查所在国家的相关法律法规。'"**
数据集概述
基本信息
- 名称: Project Gutenberg eBooks in English
- 语言: 英语 (en)
- 任务类别: 文本生成 (text-generation)
- 许可证: MIT
- 标签:
- Project Gutenberg
- e-book
- gutenberg.org
数据集结构
- 特征:
- TEXT: 文本内容,字符串类型
- SOURCE: 来源信息,字符串类型
- METADATA: 元数据信息,字符串类型
- 分割:
- train: 包含48284个样本,总大小为18104255935字节
- 下载大小: 10748877194字节
- 数据集大小: 18104255935字节
元数据详情
- METADATA 列包含的详细信息:
- language: 语言
- text_id: 唯一书籍标识符,整数类型
- title: 书籍标题,字符串类型
- issued: 发行日期,字符串类型
- authors: 作者,字符串类型,有时包含日期,逗号分隔
- subjects: 主题,字符串类型,格式多样
- locc: 美国国会图书馆分类代码,字符串类型
- bookshelves: 书架分类,字符串类型,可选
数据生成方式
- 使用爬虫从Gutenberg目录下载原始HTML代码。
- 通过额外的解析器尝试分离元数据和文本主体,并移除转录者笔记和电子书相关信息。
- 清洗后的文本主体和目录元数据保存为parquet文件,所有列均为字符串类型。
版权注意事项
- 部分书籍受版权保护,爬虫通过正则表达式忽略所有带有英文版权头的书籍。
- 使用前需手动检查每本书的元数据,确保符合使用国的法律规定。



