Developing a Specialized Question Answering Model for the Yemeni Middle School Curriculum Using NLP Techniques

Question Answering (QA) is a critical subfield in Natural Language Processing (NLP). However, research efforts have predominantly focused on the English language, leaving other linguistically rich languages, such as Arabic, significantly underexplored. Recent research has increasingly focused on developing QA systems for Islamic-Arabic texts. However, these texts present unique computational challenges due to Arabic’s complex morphology and syntax, as well as the lack of large-scale, annotated benchmark corpora. Moreover, recent studies have emphasized advancing open-domain QA frameworks that can dynamically retrieve and generate accurate, context-aware responses from both structured and unstructured knowledge sources. In this study, we employed artificial intelligence to answer questions from the Yemeni middle school curriculum (grades 7-9) to facilitate access to precise educational information. We created a textbook-derived dataset consisting of questions and answers to help students obtain accurate information without navigating overwhelming online content. We used the transformer-based Arabic language generation model AraGPT2, and we reconfigured the answer generation task, as our approach provides answers to questions by generating the answer from the trained data without relying on external contexts. This is what is known as the system for answering closed-domain questions. Then we evaluated the resulting model using the evaluation methods, which are BERTScore, Levenshtein distance, and BLEU. The proposed model achieved the best evaluation score of 0.74 on the BERTScore scale, which indicates good performance. In addition, the performance results were analyzed to determine the strengths and weaknesses of the model. The researchers suggest improving the model's performance by expanding the dataset, modifying the model structure, and applying human evaluation of the model.