Decomposing interactions

In UML 2.0 sequence diagrams have been considerably extended and are now fundamentally better structured. Interactions in sequence diagrams can be structured using so-called interaction fragments, including alt (alternative behaviour), par (parallel behaviour), neg (forbidden behaviour), assert (mandatory behaviour) and ref (reference another diagram). The operator ref in particular greatly improves the way diagrams can be decomposed. In previous work we have given a semantics to a subset of sequence diagrams using labelled event structures, a true-concurrent model that naturally captures alternative and parallel behaviour. In this paper, we expand that work to address refinement and show how to obtain a refined model by means of a powerful categorical construction over two categories of labelled event structures. The underlying motivation for this work is reasoning and verification of complex scenario-based inter-object behavioural models. We conclude the paper with a discussion on future work.