Nutritional supplements market is increasing worldwide. In 2015, the global supplements market was €109 billion and is expected to grow to €160 billion by 2020. More and more persons need help managing their diets and nutritional intake. The use of technology can increase nutritional education and improve the security of the treatment. Diet-related health anomalies are becoming more frequent not only in developing countries but also in the westernized society.
In order to address this problem, a system that interacts with the user, via a mobile application, and recommends nutritional supplements was created. This system analyzes the supplement’s label statements, and automatically extracts information regarding the effects of the supplement on the user’s health.
This system makes use of the Dietary Supplement Label Database. This dataset is composed of 65,684 different supplements, that are used for recommendation generation. For a set of 50 supplements and their respective recommendations, with the very low threshold of 10%, the system achieved a precision value of 0,39; recall value of 1 and f1-score of 0,57.
From the results it is possible to conclude that this system has some limitations regarding the extraction of relevant information from the supplement statements. It is also possible to infer that the usage of more structured information can increase the performance of such extraction.
Author: Nuno Castro
Type: MSc thesis
Partner: Faculdade de Engenharia da Universidade do Porto