A Fine-Tuned GPT-4-Based Question Answering System for E-Government Services Using a Custom-Built Dataset

This paper presented a fine-tuned GPT-4-based Question Answering (QA) system designed to improve citizen access to government services. Leveraging a custom-built dataset of over 1,000 curated government FAQs, the system delivered more accurate and context-aware responses than traditional rule-based systems. It featured a web interface, query processor, GPT-4 inference engine, and a response validator. Evaluation using BLEU (0.68 vs. 0.51) and ROUGEL (0.74 vs. 0.60) scores, complemented by human judgment on accuracy, helpfulness, and relevance, demonstrated substantial improvements over baseline systems. Robustness testing further showed the system’s strong capability to handle informal, abbreviated, and indirect queries across multiple public service domains—including civil registration, social welfare, healthcare, taxation, and employment—highlighting its potential as a scalable, user-friendly tool for digital public service delivery.