1、 通过维基百科API获取相关信息,然后依次对里面的子网页进行抓取,默认深度是4。
(Information was retrieved via the Wikipedia API, followed by crawling subpages with a default depth of 4.)
2、 爬取完成后,对内容进行粗清洗,去掉了文本中的 “参考来源” 以及 “注释” 相关内容,并把繁体中文转换为简体中文,最后把正文按句切分,方便后续发送给llm生成相关问答对。
(After crawling, a preliminary cleaning of the content was performed, removing… See the full description on the dataset page:
https://huggingface.co/datasets/Chal1ce/agricultrue_wiki_reasoning_QA_110K.