Question Answering System Based on Transformer Architecture Used to Interrogate Educational Resources in an Arabic E-Learning System

The Question-Answering (QA) task has been well-established in the Natural Language Processing (NLP) community for many years. Generally, it involves answering given questions using documents or conversations, utilizing knowledge as required and implementing inference mechanisms. In recent years, numerous machine learning algorithms have been employed to address question answering using natural language. In this paper, we propose an ensemble learning model based on Arabic variants of BERT models. Our QA system demonstrates superior results, with response accuracy exceeding 73%. The second part of the paper focuses on leveraging this high performance by integrating a prototype that concretizes the QA system as a tool to assist learners in satisfying user needs concerning response accuracy and justification. To validate our system, an experiment was conducted with a community of students using a corpus of educational resources (Electronic books from a library). Most users (94%) received correct answers to their questions, confirming the effectiveness of our system in terms of accuracy and response time.