RaGoÛt: An Arpeggio of Tastes

This paper presents the RaGoÛt system developed in response to the ECCBR’08 Computer Cooking Contest call. It proposes a case authoring approach that combines domain-specific and independent knowledge sources to create a feature vector representation of recipes. Case retrieval uses semantic similarity knowledge acquired from WordNet. This is combined with exact matching to enforce adaptation-aware case ranking, followed by substitutional adaptation of mismatched query ingredients. RaGoÛt addresses all outlined CCC tasks: compulsory, negation and menu challenges. Preliminary results from the compulsory and negation tasks are very favourable for both retrieval and adaptation. Although RaGoÛt can also retrieve three-course menus in response to the third menu task, its adaptation strategy for menus is being improved.