Building an AI-Powered Robofest Chatbot: A Fine-Tuning Approach with OpenAI and Hugging Face Exploration
Igri Fishta, BS Computer Science Candidate, CoAS; Annalia Schoenherr, BS Computer Science Candidate, CoAS; CJ Chung, PhD, Professor
Computer Science
College of Arts and Sciences, Lawrence Technological University
chatbot, OpenAI, Robofest, fine-tuning
submitted by igrifishta
This project designs and implements an AI-powered chatbot to answer questions about the Lawrence Technological University Robofest [6] event. Currently, no chatbot exists for this purpose, and creating one will enhance user-friendliness, provide valuable event information, and boost engagement and participation. The chatbot uses a fine-tuned OpenAI language model to manage both simple and complex interactions with quick and accurate responses. Key project steps include collecting and preparing JSON data on Robofest, fine-tuning the model with OpenAI and Hugging Face, and testing its performance with 20 event-related questions. The project also involves designing a user-friendly front-end using JavaScript and HTML. The final outcome was an operational chatbot that answers user questions effectively, despite limitations like the cost of running the models and challenges in dataset conversion. Iterative testing and data refinement enhanced the model's robustness, and future efforts may focus on deploying the chatbot on the live Robofest website. However, using the fine-tuning technique did not turn out very optimal because of the large amount of data updating it required, so using a different approach instead can give better results.
CC BY
Back to Search