Bilingual chatbot development for hospital UTAR using transformer

Chin, Owen (2024) Bilingual chatbot development for hospital UTAR using transformer. Final Year Project, UTAR.

Preview

PDF Download (4Mb) | Preview

Abstract

This proposal introduces a project which aimed at enhancing the user experience on UTAR Hospital's website by incorporating an innovative English Chinese multilingual chatbot. The chatbot leverages advanced technologies such as transformers, similarity search from a vector database, and natural language processing (NLP) which focused on delivering information about Traditional Chinese Medicine (TCM). The core of the project revolves around the utilization of transformers which enable the chatbot to understand and generate contextually relevant responses in both English and Chinese. This ensures seamless communication with a diverse range of website visitors which cater to the linguistic preferences of a multicultural audience. Additionally, the project integrates a similarity search mechanism using a vector database which is also known as Retrieval Augmented Generation to enhance the chatbot's ability to retrieve and present relevant TCM information. This feature provides users with personalized and accurate responses by analysing similarities between user queries and the database with TCM knowledge as well as frequently asked questions. The project's technological framework also embraces NLP to allow the chatbot to interpret and respond to user inquiries in a natural and human-like manner. This feature enhances the overall user engagement and accessibility of TCM information, fostering a user-friendly and informative experience for all website visitors. The project also aims to provide a multilingual chatbot which serve as a proficient and reliable health chat. It delivers valuable insights on Traditional Chinese Medicine by addressing the language gaps. This proposal outlines the technical architecture, implementation plan, and anticipated benefits of this project for a diverse user base.

Actions (login required)