UNI – A Retrieval Augmented Generation powered Virtual Assistant for college related queries
Sanjana J.*1, Mahadev Prasad Y. N. 2, Srinivas B. 3, Sharanya S. 4, Madhusudhan M. 5
1,2,3,4,5 Department of Information Science and Engineering, MIT Tha.ndavapura, India
Emails: sanjana.4mn20is027@gmail.com; mahadevaprasadyn@gmail.com; srinivasb20016@gmail.com; sharanyashivanna121@gmail.com; madhu1435sudhan@gmail.com
Abstract
This paper unveils an advanced chatbot engineered to cater specifically to college-related inquiries. Harnessing the power of BARD and incorporating a wake word activation system with automatic speech recognition, the chatbot offers an enhanced user experience marked by both linguistic sophistication and spoken command initiation. The methodology encompasses the nuanced process of pre-training on diverse corpora, fine-tuning to optimize responsiveness to college-specific queries, and the seamless integration of intent classification and entity recognition. These facets collectively empower the chatbot to understand and respond effectively to the intricacies of user inputs. A comprehensive knowledge base is curated to ensure not only accurate information retrieval but also to foster a depth of contextual understanding. This project signifies a pioneering leap in providing an innovative, user-friendly, and ethically driven solution for addressing college-related queries through natural language interactions. By showcasing practical advancements in chatbot technology tailored to the educational landscape, this research contributes to the evolving landscape of intelligent virtual assistants.
Keywords: virtual assistant; large language model; natural language queries; entity recognition; wake word activation; intent classification; retrieval augmented generation