Tevatron/msmarco-passage-aug

MS MARCO Passage Augmented 是一个用于段落检索的数据集，基于MS MARCO（机器阅读理解数据集）构建。该数据集专门设计用于信息检索任务，特别是段落级别的检索，旨在帮助训练和评估检索模型。它包含查询（query）及其对应的正例段落（positive_passages，即相关段落）和负例段落（negative_passages，即不相关段落），以支持监督学习。数据集特征包括查询ID（query_id，用于唯一标识查询）、查询文本（query，用户搜索的查询语句）、正例段落（每个段落包含文档ID、标题和文本）和负例段落（结构类似）。数据集适用于文档检索和段落检索任务，规模在100万到1000万之间，语言为英语。它常用于训练检索模型，如双编码器或交叉编码器，以改善搜索引擎和问答系统的性能。

MS MARCO Passage Augmented is a passage retrieval dataset based on MS MARCO (Machine Reading Comprehension Dataset). It is specifically designed for information retrieval tasks, particularly at the passage level, to aid in training and evaluating retrieval models. The dataset includes queries along with positive passages (relevant passages) and negative passages (non-relevant passages), supporting supervised learning. Features consist of query_id (unique identifier for the query), query (the search query text), positive_passages (each containing docid, title, and text), and negative_passages (similarly structured). It is suitable for document retrieval and passage retrieval tasks, with a size between 1M and 10M, and is in English. It is commonly used to train retrieval models, such as dual encoders or cross-encoders, to enhance the performance of search engines and question-answering systems.