NutriSem: A Semantics-Driven Approach to Calculating Nutritional Value of Recipes

The proliferation of recipes and other food information on the Web makes it difficult to identify recipes and foods for people who want to eat healthily. There are various tools related to calculating recipes nutritional values (NVs) but often the results obtained for the same recipe are varied. In this article we present NutriSem, a framework which allows automating the nutritional qualification of cooking recipes. It consists of four steps: (i) lexical enrichment of terms denoting ingredients; (ii) generating of nutritional calculus from lexical pattern and composition table requests; (iii) calculation and allocation of the final score; (iv) translation of the calculated score into a graphical scale. The core of the approach is based on mappings established between text corpora (cooking recipes) and structured data (food composition tables). A Knowledge Graph resource is used to enhance the quality of the mappings and therefore allow a better nutritional qualification of a given recipe.