Content adaptation on the Web reduces available information to a subset that matches a user's anticipated needs. Recommender systems rely on relevance scores for individual content items; in particular, pattern-based recommendation exploits co-occurrences of items in user sessions to ground any guesses about relevancy. To enhance the discovered patterns' quality, the authors propose using metadata about the content that they assume is stored in a domain ontology. Their approach comprises a dedicated pattern space built on top of the ontology, navigation primitives, mining methods, and recommendation techniques.