MKQA (Multilingual Knowledge Questions and Answers)

多语言知识问答 (MKQA) 是一个开放域问答评估集，包含 10k 个问答对，这些问答对排列在 26 种不同类型的语言中（总共 260k 个问答对）。该数据集的目标是为各种语言的问答质量提供具有挑战性的基准。答案基于与语言无关的数据表示，使结果可以跨语言进行比较，并且独立于特定语言的段落。该数据集包含 26 种语言，为评估问答提供了迄今为止最广泛的语言。

Multilingual Knowledge Question Answering (MKQA) is an open-domain question answering evaluation benchmark consisting of 10,000 question-answer pairs per language across 26 distinct languages, totaling 260,000 question-answer pairs. The core objective of this dataset is to provide a challenging benchmark for assessing question answering quality across diverse languages. Answers are grounded in language-agnostic data representations, enabling cross-linguistic comparison of evaluation results and independence from language-specific source passages. Covering 26 languages, this dataset offers the broadest linguistic coverage for question answering evaluation to date.

搜集汇总

背景与挑战

背景概述

MKQA是一个多语言开放域问答评估数据集，包含10k个问答对，涵盖26种语言，旨在为多语言问答质量提供基准测试。该数据集由苹果公司于2020年发布，支持跨语言比较，具有广泛的语言覆盖范围。