Semantic food knowledge model

In order to reflect local food culture, LIFANA uses country-specific recipe datasets with precise information about the nutrients.​

Cooking recipes usually consist of instructions, a list of ingredients, and quantities. We apply Natural Language Processing (NLP) tools to parse them and match the ingredients with foods indexed in the Food Composition Databases (FCDBs) containing (macro- and micro-) nutritional information.​

Some internationally known recipes can be used in all countries but must be localized (translated and linked to the national FCDB) because the composition of similar food products differs because of the climate, the production processes, and other factors. We align FCDBs by comparing food characteristics, such as the source, packaging, and treatment.​