Desarrollo de un sistema integrado de captura, procesamiento y consulta de información nutricional de productos procesados y ultra procesados: una aplicación móvil con asistente virtual basado en RAG

Abstract

The increasing consumption of processed and ultra-processed foods poses significant public health challenges due to its association with non-communicable diseases. This study proposes an integrated system for capturing, processing, and querying nutritional information of these products, specifically tailored for the Ecuadorian context. The system includes a mobile application that utilizes Optical Character Recognition (OCR) and a Named Entity Recognition (NER) model to extract data from product labels, which are then stored in a knowledge graph managed with Neo4j. A web platform enables the validation and management of the captured data, while a virtual assistant based on Retrieval-Augmented Generation (RAG) facilitates natural language queries, generating precise and context-aware responses. The system architecture combines advanced technologies such as Google Vision API, OpenAI, Flutter, and Angular, providing an efficient solution adapted to local needs. Preliminary evaluations demonstrate high levels of usability and accuracy, showing that the system not only automates the collection of nutritional information but also enhances its accessibility and reliability, significantly contributing to data management in this field.