Usage of data science techniques to personalize and optimize nutrition recommendations and information via a micronutrient focused application
Author
Term
4. semester
Publication year
2024
Submitted on
2024-06-03
Pages
64
Abstract
This paper explores the usage of data science techniques, like retrieval augmented generation chatbots, graphRAG, object detection machine learning models, and gen- erally LLMs as well as speech-to-text to create a new nutrition application. This new product shall be able to offer a new value proposition and thus, differentiate itself from preexisting solutions. After regarding prior research in the field, publish- ing and analyzing a survey of potential users, and talking to an actual nutritionist, a difference can be drawn between an aesthetical fitness and a health focus in applica- tions. While most products in the market concentrate on the first one, this paper’s application represents a health- and micronutrient-focused solution. The central element is a food recommendation system based on the concept of vector similar- ity. This way, the application is not only a food tracker, but actually offers more functionality, namely the recommendation of foods based on their nutrient profile. Interpreted from the survey, the logging of foods was furthermore perceived as a pain point in usage of other applications. Hence, computer vision and speech-to-text was successfully used to offer an alternative to the slower, manual process of typing in food names. Throughout, LLMs are a central technology to quickly implement new artificial intelligence-based functionalities, whether solely text-based or multimodal.
Documents