EduQAS: Arabic Baseline Dataset to Advance Question Answering in Moroccan Universities

Abstract

Arabic question-answering systems are essential tools that provide rapid access to relevant information in the Arabic language, particularly in specialized domains such as healthcare, education, and beyond. This study presents the development of a novel Arabic question-answering dataset focused on the Moroccan university context, serving as a baseline for improving Arabic question-answering systems and representing a key step in an underexplored domain where no dedicated resources previously existed. To address this gap, we introduce EduQAS, a dataset aimed at answering university-related queries tailored to the needs of Moroccan students. EduQAS consists of diverse and reliable sources, including official university websites, legislative documents, and information provided by relevant ministries. Beyond serving as a knowledge base, EduQAS intended to support the training and evaluation of state-of-the-art question-answering models. In this work, three models-AraT5, AraBERTv2, and mT5-small-were trained on the dataset. The experimental results demonstrate strong performance, with high scores comparable to those obtained on existing question-answering datasets.