alexandrainst/foqa

FoQA是一个法罗语提取式问答（也称为阅读理解）数据集，包含2000个问题-答案-上下文三元组，上下文来自法罗语维基百科文章。数据集通过两阶段过程创建：首先，使用GPT-4-turbo自动生成10000个问题-答案-上下文三元组，然后由一位母语法罗语使用者对这些三元组进行人工审查，最终得到2000个三元组。所有数据点均可用，包括被拒绝或未经验证的数据点。

FoQA is a Faroese extractive question answering (also known as reading comprehension) dataset, consisting of 2,000 question-answer-context triples, with the contexts coming from Faroese Wikipedia articles. The dataset has been created through a two-stage process: First, 10,000 question-answer-context triples were automatically generated using GPT-4-turbo, and then manually reviewed by a native Faroese speaker to result in the final 2,000 triples. All data points are available, including the ones that were rejected or not manually validated.