Using SemWeb/Ontology to improve Machine Learning results

Does anybody know of work done to improve the results of machine learning on a set of data by first giving the data a semantic treatment?

Most papers that Google shows me use machine learning techniques to help perform semweb tasks like semantic annotation or ontology generation. Much of this work is very cool, but I'm interested in the opposite direction. For example, ways that a machine learning algorithm for Netflix recommendations be improved with ontologizing.

You can have a look at the work by the Rebholz group at EBI. Who use ontologies to improve text mining results. I suspect there are more out there but its outside my area of expertise.