Design and implementation of a health-oriented dietary recommendation system based on natural language understanding

Summarise with VIRAT
Authors: Zhizhang Mei
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This dissertation proposes and implements a health-oriented meal recommendation system driven by natural language understanding. It is motivated by the increasing demand for balanced diets and chronic disease prevention, particularly in food-rich regions like Singapore. The system bridges personal dietary preferences with data-driven nutritional guidance. Built on LangGraph orchestration, it integrates large language model (LLM) intent parsing, restaurant search via Google Maps API, menu extraction through Tavily, and fuzzy alignment with nutrition databases such as USDA and HPB. The system processes natural language queries end-to-end and generates ranked meal recommendations with structured nutrition summaries and explainable health assessments. Real-world testing at Geylang Serai Market achieved over 90\% health compliance in the top three results. These experiments demonstrate consistent reasoning across diverse dietary constraints. A layered SQLite caching mechanism improves efficiency and achieves a 70\% hit rate. It also reduces LLM calls by more than half. While data sparsity and limited labeling remain challenges, the framework is extensible. Future work will explore vector-based nutrition retrieval and multimodal extensions such as meal photo recognition. Overall, this study provides a practical and interpretable foundation for intelligent dietary recommendation systems in real-world health contexts.

Fields of Study & Subjects

Innovation in Digital Healthcare Systems Recommender Systems and Techniques Information Retrieval and Search Behavior

Publication details

Journal / source
DR-NTU (Nanyang Technological University)
Publication date
1 January 2026
Language
EN
Document type
Thesis