Enhancing Content-Based Recommendation with the Task Model of Classification

In this paper, we define reusable inference steps (realize, classification by concepts, classification by instances and retrieve) for content-based recommender systems applied on semantically-enriched collections. In our case, we use the enriched museum collection. The core steps: (i) Classification by concepts brings explicitly related concepts via artwork features and semantic relations between artworks and concepts, e.g. “The Night Watch” has creator “Rembrandt van Rijn” and “Rembrandt van Rijn” is a student of “Pieter Lastman”; and (ii) Classification by instances brings implicitly related concepts using the method of instance-based ontology matching, e.g. “Cupid” is implicitly related to “Love and sex” because they describe sufficient artworks in common. To combine predictions from these two steps for each related concept, we set a parameter α to balance the strength of explicit and implicit recommendations. We test our strategy with the CHIP Art Recommender in terms of accuracy and discuss the added values of providing serendipitous recommendations and supporting more complete explanations for