KUCHING OLD BAZAAR QUESTION AND ANSWER SYSTEM THROUGH AI ENABLED WHATSAPP
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UNIVERSITI MALAYSIA SARAWAK
Abstract
Description
Kuching Old Bazaar, a 200-year-old Chinatown and former trading hub from the Brooke era now offers a vibrant and mix historical charm and modern goods, including handicrafts and local produce, while being divided into the Old Bazaar by the Sarawak River and the New Bazaar near India Street. Despite its rich historical significance, many are unaware of the history this historical site prevails. This is most likely due to limited information, slow and inefficient retrieval from online sources. Other problems include scattered information from various sources and too many applications needed to find them. In this paper, a question-and-answer system through AI enabled WhatsApp was proposed and designed. It allows general users to communicate with the chatbot regarding Kuching Old Bazaar through WhatsApp without needing to download additional applications. Based on user requirements survey, 35 out of 40 respondents (87.5%) would want to use WhatsApp to learn about historical places like Kuching Old Bazaar. The system is a hybrid combination of Graph Retrieval-Augmented Generation (RAG) technique using Neo4j knowledge graph and fine-tuned Flan T5 (Encoder-Decoder) model to understand user query and reply with relevant context and continuity. From the testing and evaluation, this hybrid combination boasts a higher percentage (96.5%) at displaying accurate and relevant information to the user compared to only using the encoder-decoder model (81%) and knowledge graph (85%) independently. Further research includes voice and image recognition, lesser dependencies on Gemini API call and propose a distributed chatbot system.