Extracting Vegetable Information from Recipes to Facilitate Health-Aware Choices

In order to facilitate users of a recipe service to make health-aware meal choices, we focus on the automatic extraction of nutritional information from the text of a recipe. More specifically, the presence of sufficient vegetables in the main meal is addressed in this note. We present an algorithm to extract this information from the recipe texts and explore its use to persuade users to select vegetable-rich meals. The consumption of a sufficient amount of vegetables is a prominent, universal health recommendation. Moreover, vegetable rich menus are typically low on the number of calories and promote satiety [3]. A 2003 study [2] shows that the average consumption of vegetables in the Netherlands is lagging far behind the generally known governmental recommendation of 200 grams a day. Obviously, people who are motivated to select a vegetable-rich recipe can inspect the ingredient lists themselves. The mere presence of vegetables in a recipe is fairly easy to recognize, but the quantities of vegetables in terms of grams may be more problematic. For example, when an ingredient list contains entries like two large onions, a can of chopped tomatoes and 1/2 leek, it is not straight forward to translate such pieces and units to grams. By making the presence of sufficient vegetables in a meal salient, health-aware choices can be facilitated.