shomez/blink-pretrain-wikipedia-flat

该数据集包含文本输入及其相关元数据，用于某种形式的文本分析或实体链接任务。每个样本包括一个输入字符串、元数据（包含提及、左上下文和右上下文）、维基百科ID标签和标题标签。数据集包含900万个训练样本，总大小约为9.75GB。

This dataset contains text inputs and associated metadata, likely for some form of text analysis or entity linking task. Each sample includes an input string, metadata (containing mention, left context, and right context), a Wikipedia ID label, and a title label. The dataset includes 9 million training samples, with a total size of approximately 9.75GB.