Corpus Construction for Automatic Meal Planning

Eating a healthy diet every day is ideal, but difficult for many people. There is a scarcity of information and available resources regarding ideal food combinations, limiting the possibility of developing effective solutions for culinary information processing. As a result, food recommendations based on existing data tend to be unhealthy because the majority of users lack nutritional knowledge. While collecting healthy and tasty menus through crowdsourcing is one solution, such data-input tasks must be carefully designed from a combinatorial perspective. We describe a methodology to operationalize the collection of appropriate data for this complex task, present notable features of the resulting data resource, and illustrate how the obtained data can be referenced in automatic menu planning.