A typical model merging session: requires a great deal of knowledgeable input; does not provide rapid feedback; quickly overwhelms the user with details; fails to properly match elements; performs minimal conflict detection; offers conflict resolution choices that are inadequate and without semantics; and exhibits counter-intuitive behavior.

Viewing model merging as a process, this research defines a hybrid merge workflow that blends the best of the main approaches to merging, expressing its phases as algebraic operators for performing transformations on model and relationship data types. Normalization and denormalization phases decouple models from their originating tool and metamodel. State-based phases capture model differences in the model itself, establish element correspondence using multiply matching strategies, and extract change operations. Operation-based phases then partition and order the changes prior to the detection and automatic resolution of conflicts.

The work has culminated in a prototype that validates the workflow, while realizing several novel model merging ideas, which are evaluated with simple and involved test cases. Combining the hybrid merge approach with the semantic expressiveness of decision tables---open to user modification---and an interactive and batch mode of operation allows the tool, named Mirador, to successfully address, to varying degrees, all of the previously cited shortcomings.