HawkerSearch: an LLM-powered marketing platform for elderly and low-tech literacy hawkers in Singapore

• Main Author: Yap, Shawn Yu Xiang
• Other Authors: Anwitaman Datta
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Hawker; Marketing
• Online Access: https://hdl.handle.net/10356/181157
• Institution: Nanyang Technological University

This project presents the design, development, and testing of HawkerSearch, a mobile application powered by Large Language Models (LLMs) to deliver personalised hawker stall recommendations. By matching hawker stalls to user preferences, the app reduces decision paralysis—a common problem when choosing food options—and highlights the healthy food choices available at hawker centres, challenging the misconception that hawker food is generally unhealthy. Additionally, the app provides hawker stall owners with LLM-driven business analytics and insights to help them optimize their offerings and attract more customers. The core innovation lies in the novel tiered LLM-powered recommendation system, which integrates both traditional recommendation techniques and LLM features to ensure high accuracy and scalability across Singapore’s large number of hawker stalls. This approach overcomes the limitations of existing LLM-based recommendation systems, which often lack precision or fail to scale effectively. Testing with real hawker stall owners and customers yielded promising results. The app reduced decision paralysis by 50%, and the recommendation system received an average relevance score of 7/10. Furthermore, interviews with hawker stall owners revealed that the analytics page was rated 7/10 in usefulness, demonstrating the app’s potential to support hawker businesses by providing valuable insights into customer preferences and operational improvements.