Making sense to modelers: Presenting UML class model differences in prose

Understanding the difference between two models, such as different versions of a design, can be difficult. It is a commonly held belief in the model differencing community that the best way of presenting a model difference is by using graph or tree-based visualizations. We disagree and present an alternative approach where sets of low-level model differences are abstracted into high-level model differences that lend themselves to being presented textually.

This format is informed by an explorative survey to elicit the change descriptions modelers use themselves. Our approach is validated by a controlled experiment that tests three alternatives to presenting model differences. Our findings support our claim that the approach presented here is superior to EMF Compare.