WebFAQ

该数据集是一个大规模的开域问答数据集，来源于FAQ风格的schema.org注释，包含了9600万个自然语言问答对，覆盖了75种语言。其中，非英语样本达4700万个，并为20个单语检索基准提供了基础数据。该数据集已用于微调领域内的预训练XLM-RoBERTa模型，显著提升了检索性能。其规模达到了9600万问答对，任务涵盖了多语言问答和检索。

This is a large-scale open-domain question answering dataset derived from FAQ-style schema.org annotations. It contains 96 million natural language question-answer pairs across 75 languages, of which 47 million are non-English samples. This dataset serves as the foundational resource for 20 monolingual retrieval benchmarks, and has been utilized to fine-tune the pre-trained XLM-RoBERTa model for the question answering domain, achieving significant improvements in retrieval performance. With a total scale of 96 million question-answer pairs, its supported tasks cover multilingual question answering and retrieval.