From Multiple-Choice to Extractive QA: A Case Study for English and Arabic

TL;DR

This work investigates whether a multilingual MCQA dataset (Belebele) can be repurposed into an extractive QA (EQA) resource for English, Modern Standard Arabic, and Arabic dialects. It develops Belebele-EQA by manual span annotation guided by language-specific rules, with a bootstrapped MSA alignment and a parallel English–MSA dataset, complemented by a dialect QA benchmark. Using PrimeQA with pretrained multilingual models, the study shows that EQA from MCQA is feasible but requires substantial manual curation; results reveal lower performance on Belebele-EQA than on SQuAD, with translation and subset filtering improving cross-lingual and dialectal performance. The work provides annotation guidelines, IAA evidence, and practical insights to broadly extend EQA resources to the 120 Belebele languages, supporting faster development of QA for resource-poor languages and informing future automation efforts.

Abstract

The rapid evolution of Natural Language Processing (NLP) has favoured major languages such as English, leaving a significant gap for many others due to limited resources. This is especially evident in the context of data annotation, a task whose importance cannot be underestimated, but which is time-consuming and costly. Thus, any dataset for resource-poor languages is precious, in particular when it is task-specific. Here, we explore the feasibility of repurposing an existing multilingual dataset for a new NLP task: we repurpose a subset of the BELEBELE dataset (Bandarkar et al., 2023), which was designed for multiple-choice question answering (MCQA), to enable the more practical task of extractive QA (EQA) in the style of machine reading comprehension. We present annotation guidelines and a parallel EQA dataset for English and Modern Standard Arabic (MSA). We also present QA evaluation results for several monolingual and cross-lingual QA pairs including English, MSA, and five Arabic dialects. We aim to help others adapt our approach for the remaining 120 BELEBELE language variants, many of which are deemed under-resourced. We also provide a thorough analysis and share insights to deepen understanding of the challenges and opportunities in NLP task reformulation.

Figures (1)

• Figure 1: Visualization of the IAA agreement for 42 questions. The colored lines indicate a span. Each color represents an annotator. The number on the colored line is a question ID.