This paper discusses the integration of ontologies with service choreographies in view of recommending interest points to the modeler for model improvement. The concept is based on an ontology of recommendations (evaluated by metrics) attached to the elements of the model. The ontology and an associated knowledge base are used in order to extract correct recommendations (specified as textual annotations attached to the model) and present them to the modeler. Recommendations may result in model improvements. The recommendations rely on similarity measures between the captured modeler design intention and the knowledge stored in the ontology and knowledge bases.