NutriRAG: Unleashing the Power of Large Language Models for Food Identification and Classification through Retrieval Methods.

Abstract

Objective: This study explores the use of advanced Natural Language Processing (NLP) techniques to enhance food classification and dietary analysis using raw text input from a diet tracking app.

Materials and methods: The study was conducted in three stages: data collection, framework development, and application. Data were collected via the myCircadianClock app, where participants logged their meals in free-text format. Only de-identified food-related entries were used. We developed the NutriRAG framework, an NLP framework utilizing a Retrieval-Augmented Generation (RAG) approach to retrieve examples and incorporating large language models such as GPT-4 and Llama-2-70b. NutriRAG was designed to identify and classify user-recorded food items into predefined categories and analyzed dietary patterns from free-text entries in a 12-week randomized clinical trial (RCT: NCT04259632 ). The RCT compared three groups of obese participants: those following time-restricted eating (TRE, 8-hour eating window), caloric restriction (CR, 15% reduction), and unrestricted eating (UR).

Results: NutriRAG significantly enhanced classification accuracy and effectively identified nutritional content and analyzed dietary patterns, as noted by the retrieval-augmented GPT-4 model achieving a Micro F1 score of 82.24. Both interventions showed dietary alterations: CR participants ate fewer snacks and sugary foods, while TRE participants reduced nighttime eating.

Conclusion: By using AI, NutriRAG marks a substantial advancement in food classification and dietary analysis of nutritional assessments. The findings highlight NLP's potential to personalize nutrition and manage diet-related health issues, suggesting further research to expand these models for wider use.