Testing complex Computer-Based Systems is not only a demanding but a very critical task. Therefore the use of models for generating test data is an important goal. Tool support during the generation of test cases can considerably reduce the effort and the risk of errors of this task. While model the understanding of model transformation got better in the past, the analysis of the input — requirement specifications mostly consisting of natural language texts — still represents a bottleneck. In this paper a method for closing the gaps between manual techniques for structuring texts and automatic techniques based on linguistics is explained. By providing suggestions for missing or ambiguous terms the method supports the formalization within large projects. The suggestions are derived from a glossary and from an analysis of their integration in use case descriptions. Additionally the traceability links via feature models and other model elements are analyzed. Thus, the advantages of both techniques are joined to increase the level of tool support, resulting in a higher efficiency of the test case generation.