An AI-Powered Chatbot Using a Transformer-Based Language Model for Graduate Educational Schemes Recommendation
摘要
Availability of higher education is still the priority concern for Indian socio-economically underprivileged and rural students at 25%, according to the National University of Educational Planning and Administration (NUEPA) 2021–2022 annual report. The major hurdles are financial constraints and unavailability of information regarding government-funded education programs. An AI-based learning advice chatbot leveraging the Large Language Model Meta AI (LLaMA) has been proposed to address the issue. The model is founded on the LLaMA 3.1 transformer model specifically optimized for low-resource environments. The model was fine-tuned on a 1,200 proprietary educational dataset of scholarship and government program-oriented queries with a test accuracy of 94.39% with an 80%−20% train-test split. The 60–40% was employed as another split to verify its performance with the test accuracy at 92.97%. The chatbot provides user-specific recommendations for 24/7 availability, ease, and strong data security. Providing timely and accurate information will help reduce the financial and informational barrier, thereby increasing the equitable opportunity for people from low-income backgrounds and the community to access higher education.