TyDi QA: A Benchmark for Information-Seeking Question Answering in Typologically Diverse Languages

TL;DR

TyDi QA introduces a challenging, multilingual question answering benchmark across 11 typologically diverse languages, enabling evaluation of models' generalization capabilities in real-world, language-diverse scenarios.

Contribution

The paper presents a new multilingual QA dataset covering 11 diverse languages, with high-quality, naturally collected questions to evaluate cross-lingual and typological generalization.

Findings

Dataset contains 204K question-answer pairs.

Languages exhibit diverse linguistic features.

Questions are naturally collected without translation.

Abstract

Confidently making progress on multilingual modeling requires challenging, trustworthy evaluations. We present TyDi QA---a question answering dataset covering 11 typologically diverse languages with 204K question-answer pairs. The languages of TyDi QA are diverse with regard to their typology---the set of linguistic features each language expresses---such that we expect models performing well on this set to generalize across a large number of the world's languages. We present a quantitative analysis of the data quality and example-level qualitative linguistic analyses of observed language phenomena that would not be found in English-only corpora. To provide a realistic information-seeking task and avoid priming effects, questions are written by people who want to know the answer, but don't know the answer yet, and the data is collected directly in each language without the use of translation.