MS MARCO: A Human Generated MAchine Reading COmprehension Dataset

数据集

Abstract

In MS MARCO, all questions are sampled from real anonymized user queries.

• The context passages are extracted from real web documents .
• The answers to the queries are human generated.

Introduction

Compared to previous publicly available datasets, this dataset is unique in the sense that

• all questions are real user queries,
• the context passages, which answers are derived from, are extracted from real web documents,
• all the answers to the queries are human generated,
• a subset of these queries has multiple answers,
• all queries are tagged with segment information.

Related work