HenriqueGodoy/extract-0
收藏官方服务:
资源简介:
Extract-0文档信息提取数据集是一个包含280,128个合成训练示例的数据集,用于训练Extract-0语言模型,该模型在文档信息提取任务上表现优于GPT-4和其他大型模型。数据集由来自arXiv论文、PubMed Central文章、Wikipedia内容和FDA监管文档的文本片段构成,每个示例都配有一个基于模式的提取任务和相应的结构化输出。
The Extract-0 Document Information Extraction Dataset is a collection of 280,128 synthetic training examples used to train the Extract-0 language model, which outperforms GPT-4 and other large models on document extraction tasks. The dataset consists of text chunks from arXiv papers, PubMed Central articles, Wikipedia content, and FDA regulatory documents, each paired with a schema-based extraction task and its corresponding structured output.
提供机构:
HenriqueGodoy


