LETOR4.0
收藏arXiv2025-09-30 收录
下载链接：
https://www.microsoft.com/en-us/research/publication/letor-benchmark-collection-research-learning-rank-information-retrieval/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为LETOR4.0，是评估学习排序方法的基准数据集，其包含了从GOV2语料库中提取的查询-文档对。此外，该数据集采用了三级分级标注策略（0：不相关，1：相关，2：最相关）来收集真实标签。数据集规模包含69,623个查询-文档对，其任务是对这些对进行排序。
This dataset, named LETOR 4.0, is a benchmark dataset for evaluating learning-to-rank methods. It comprises query-document pairs extracted from the GOV2 corpus. Furthermore, it employs a three-level graded annotation strategy (0: irrelevant, 1: relevant, 2: most relevant) to gather ground-truth labels. The dataset contains a total of 69,623 query-document pairs, with the core task being to rank these pairs.
提供机构：
Microsoft Research
