Lama AlQarni

Artificial Intelligence Specialization · Cohort 2024

Leveraging Language Models for Islamic Question & Answering System

Recent advancements in Natural Language Processing (NLP) have significantly enhanced the capabilities of large language models (LLMs) in understanding and generating human-like text. This progress offers promising opportunities in various domains, including the development of automated question-answering systems. In the context of Islamic question answering, these systems face unique challenges due to the complexity and sensitivity of religious texts. This research explores the potential of leveraging state-of-the-art LLMs, specifically AraT5 and BART, fine-tuned on Islamic texts, to generate accurate and contextually relevant answers to questions about Islamic jurisprudence, theology, and ethics. A baseline transformer model that was built from scratch was also developed to provide a comparative evaluation of these advanced models. The fine-tuning process involved adapting AraT5, a model tailored for Arabic text, and BART, known for its robust text generation capabilities, to a dataset of Islamic fatwas.

Read the final report (PDF)

See it in the program report (PDF)