Interpreting Graphic Models

When using graphic models for system speci cation and documentation, it is important that the information presented is readily comprehensible to the intended readers. For data models, the readers are primarily the system users, who are asked to con rm the correctness of the information and data requirements modelled, and the system implementers who are to implement the database system in accordance with the speci cations given in the model. It is commonly argued that graphic models enhance understanding. The authors of graphic data models appear to assume that the graphic model is a su cient speci cation tool, in as much as it is seldom supplemented with a de nition language other than translation rules to a relational model. We have studied user comprehension of 3 di erent representation styles (1) used for graphic data models. Our measure of user comprehension is the percentage of correctly identi ed model components. Our study indicates that the graphic representation used for a model in uences the quality of interpretation. Interpretation scores for the highly graphic models in our study were signi cantly (2) lower than the interpretation scores for minimally graphic models or for models using an embedded symbol structure. An explanation could be graphic overload which hinders component identi cation. Further, our observations show that graphic models are not completely read. Up to 3010-20models alone are not dependable tools for either requirements con rmation or reliable system implementation. It has been assumed that graphic models would be read in a text-like serial manner. However, the most pro cient interpreters focused on identi cation and presentation of the model structure, given in the central portion of the models. All subjects used this interpretation strategy for the familiar model types, indicating that it can be learned.