LMGQS

LMGQS是一个大规模的查询焦点摘要数据集，由微软亚洲研究院创建。该数据集包含超过113万个文档-查询-摘要三元组，覆盖了广泛的文档和问题类型。创建过程中，研究人员利用大规模预训练语言模型InstructGPT从四个通用摘要基准中提取隐藏查询，高效地扩展了标注。LMGQS数据集的应用领域主要集中在查询焦点摘要任务，旨在解决现有QFS数据集规模不足的问题，推动QFS领域的研究发展。LMGQS is a large-scale query-focused summarization dataset created by Microsoft Research Asia. It contains over 1.13 million document-query-summary triples, covering a wide range of document and question types. During its creation, researchers leveraged the large-scale pre-trained language model InstructGPT to extract hidden queries from four generic summarization benchmarks, enabling efficient annotation scaling. The LMGQS dataset is primarily targeted at the query-focused summarization (QFS) task, aiming to address the insufficient scale of existing QFS datasets and advance research in the QFS domain.