Searching for cooking recipes presents a challenge with seemingly endless variation and little standardisation even across individual catalogues. The root cause is a total lack of consistent semantics, but there is a fix. Semantic reasoning over an ontology provides an elegant solution, enriching a knowledge graph for fast, easy, and contextual search.
The language used in recipes varies from author to author as they convey information about the ingredients, cooking style, and background. While helpful to the human reader, these can cause issues for a digital search. For example, inconsistencies such as ‘seabass’ and ‘sea bass’ become problematic. Furthermore, there is no taxonomy that tells us that seabass is a fish, or that a vegan recipe has no animal products, and of course nothing that tells us which products are animal products to begin with.
To navigate these obstacles we form an ontology which can then be used to standardise and enhance tagging of recipes. The clear hierarchy of terms lends itself to more forgiving searches while maintaining the accuracy of their meaning. New concepts are created using OWL and Datalog rules that make the search experience smoother, using negation as failure to define the notion of ‘vegan’, and aggregation to calculate calory counts in a normalised and easily maintainable way.
