This paper describes the integration of linguistic knowledge into the German prototype SUNGerm of the European SUN-DIAL project. The goal of the project is to construct a telphone based real time system for oral inquiries into a database of intercity train schedules [1]. We will describe the system set up and the interaction among the speech recognition, linguistic analyser, and dialogue modules. We suggest how these modules can effectively interact to improve recognition and understanding through language models dependent on the context of the dialogue. Additionally, we outline the linguistic analysis concerning phenomena frequently encountered in oral dialogue. We describe how these phenomena are captured in a semantic interface language (SIL) which is uniform for the different language prototypes within the SUNDIAL project.