TyDi QA (Typologically Diverse Question Answering)

TyDi QA 是一个问答数据集，涵盖 11 种类型不同的语言，包含 20 万个问答对。 TyDi QA 的语言在其类型学（每种语言表达的语言特征的集合）方面是多种多样的，因此作者希望模型在该集合上表现良好，能够泛化世界上的大量语言。

TyDi QA is a question answering dataset encompassing 11 typologically diverse languages, with a total of 200,000 question-answer pairs. The languages in this dataset vary widely in terms of linguistic typology, which refers to the collection of linguistic features each language exhibits. As such, the authors designed TyDi QA such that models achieving strong performance on it can generalize across a vast array of the world's languages.

搜集汇总

背景与挑战

背景概述

TyDi QA是一个涵盖11种类型学多样语言的问答数据集，包含20万个问答对。该数据集旨在通过其语言多样性促进模型在多种语言上的泛化能力。