UQA: Corpus for Urdu Question Answering

TL;DR

This paper presents UQA, a new Urdu question answering dataset created by translating SQuAD2.0 using the EATS method, and benchmarks multilingual QA models on it, demonstrating its utility for NLP research in low-resource languages.

Contribution

The paper introduces UQA, a high-quality Urdu QA dataset generated via EATS, and evaluates state-of-the-art models, advancing multilingual NLP for low-resource languages.

Findings

XLM-RoBERTa-XL achieved an F1 score of 85.99.

EATS effectively preserves answer spans in translation.

UQA enhances cross-lingual transfer for Urdu NLP.

Abstract

This paper introduces UQA, a novel dataset for question answering and text comprehension in Urdu, a low-resource language with over 70 million native speakers. UQA is generated by translating the Stanford Question Answering Dataset (SQuAD2.0), a large-scale English QA dataset, using a technique called EATS (Enclose to Anchor, Translate, Seek), which preserves the answer spans in the translated context paragraphs. The paper describes the process of selecting and evaluating the best translation model among two candidates: Google Translator and Seamless M4T. The paper also benchmarks several state-of-the-art multilingual QA models on UQA, including mBERT, XLM-RoBERTa, and mT5, and reports promising results. For XLM-RoBERTa-XL, we have an F1 score of 85.99 and 74.56 EM. UQA is a valuable resource for developing and testing multilingual NLP systems for Urdu and for enhancing the cross-lingual transferability of existing models. Further, the paper demonstrates the effectiveness of EATS for creating high-quality datasets for other languages and domains. The UQA dataset and the code are publicly available at www.github.com/sameearif/UQA.