A very recent topic in CBR research deals with the automated optimisation of similarity measures—a core component of each CBR application—by using machine learning techniques. In our previous work, a number of approaches to bias and guide the learning process have been proposed aiming at more stable learning results and less susceptibility to overfitting. Those methods support the learner by incorporating background knowledge into the optimisation process. In this paper, we focus on one specific form of knowledge, namely vocabulary knowledge implicitly contained in the model of the respective application domain, as a source to enhance the learning of similarity measures.