Mr. TYDI
收藏资源简介:
Mr. TYDI是一个多语言基准数据集,专为11种类型多样的语言设计,用于评估单语检索中的学习密集表示排序。该数据集由滑铁卢大学计算机科学学院创建,旨在推动非英语语言密集检索技术研究。数据集包含约5800万条数据,来源于维基百科文章,通过提取自然话语单元生成。Mr. TYDI的应用领域主要集中在非英语语言的信息检索和问答系统,旨在解决现有技术在分布外数据上表现不佳的问题。
Mr. TYDI is a multilingual benchmark dataset tailored for 11 typologically diverse languages, aimed at evaluating learned dense representation ranking in monolingual retrieval. Developed by the School of Computer Science at the University of Waterloo, this dataset is designed to advance research on dense retrieval technologies for non-English languages. It comprises approximately 58 million instances sourced from Wikipedia articles and generated via extraction of natural discourse units. The core application areas of Mr. TYDI focus on information retrieval and question answering systems for non-English languages, with the objective of addressing the underperformance of existing technologies on out-of-distribution data.

- 1Mr. TyDi: A Multi-lingual Benchmark for Dense Retrieval滑铁卢大学计算机科学学院 · 2021年



