Automating Information Retrieval from Faculty Guidelines: Designing a PDF-Driven Chatbot powered by OpenAI ChatGPT
Mutiara Auliya Khadija, Abdul Aziz, Wahyu Nurharjadmo
Abstract
Educational materials, including guides, tutorials, and master plans, are universally presented in e-book formats. This facilitates a comprehensive understanding for academics, encompassing both technical intricacies and broader conceptual frameworks. E-books offer several benefits, such as searchability and the incorporation of links to additional information sources. However, many individuals express concerns that e-books are not particularly comfortable for extended reading periods. In the other hand, a Generative AI approach is employed for the development of an intelligent chatbot. Our primary contribution lies in an automated information retrieval method, involving the design of a PDF-Driven Chatbot using Large Language Models (LLMs) in the context of faculty guidelines question answering. This research utilizes the LangChain Framework, OpenAI’s Chat-GPT (GPT3.5 Turbo), and Pinecone for generating responses. The outcomes demonstrate that the chatbot is capable of generating coherent responses closely aligned with the context of the PDF document.