Empirical studies attest that conceptual models created in distributed modelling environments often vary heavily in the way their respective model elements are labelled. Although the same issues are being modelled, different names are chosen by the involved persons. By this, the analysis and comparison of the models, which is required for their subsequent integration, is extremely challenging and time consuming. Literature analysis reveals several approaches addressing this problem by either manually or semi-automatically integrating existing models after their construction. However, this proves to be an exhaustive and error prone task. In this article we propose a domain and modelling language independent approach that prevents the emergence of naming conflicts already during the modelling process. This is done by formalising naming conventions consisting of context specific thesauri and customised phrase structures, which are both derived from natural language grammars and supplemented by domain-specific terms. These conventions serve as basis for a fully automated guidance of the modeller during the model creation process, resulting in semantically comparable conceptual models. For this, we present a research prototype that integrates our approach into a modelling tool.