A Bayesian approach to diagram matching with application to architectural models

Abstract

IT system architectures, as well as other systems, are often described by formal models or informal diagrams. In practice, there are often a number of versions of a model, e.g. for different views of a system, divergent variants, or a series of revisions. Understanding how versions of a model correspond or differ is crucial, yet little work has been done on automated assistance for matching models and diagrams.We have designed a framework based on Bayesian methods for finding these correspondences automatically. We represent models and diagrams as graphs whose nodes have attributes such as name, type, connections, and containment relations, and we have developed probabilistic models for rating the quality of candidate correspondences based on various features of the nodes in the graphs. Given the probabilistic models, we can find high quality correspondences using search algorithms. Preliminary experiments focusing on architectural models suggest that the technique is promising.

Automatic Inference of Structural Changes for Matching across Program Versions

“…Tools differ in their underlying program representation, matching granularity, matching multiplicity, and matching heuristics. In prior work [23], we compared existing matching techniques [3,5,17,18,19,20,24,26,28,32,35,37] along these dimensions. Our survey found that fine-grained matching techniques often depend on effective mappings at a higher level.…”

Behavioural model fusion

“…Many existing approaches to model merging concentrate on syntactic and structural aspects of models to identify their relationships and to combine them. For example, (Melnik, 2004) studies matching and merging of conceptual database schemata, (Mehra et al, 2005) proposes a general framework for merging visual design diagrams, describes an algebraic approach for merging requirements views, and (Mandelin et al, 2006) provides a technique for matching architecture diagrams using machine learning. These approaches treat models as graphical artifacts while largely ignoring their semantics.…”

“…There are several approaches to this, including linear and nonlinear averages, and machine learning. Learning-based techniques have been shown to be effective when proper training data is available (Mandelin et al, 2006). At this stage, we do not have sufficient training data to employ such techniques.…”

“…(Alspaugh et al, 1999) proposes term matching based on project glossaries for finding similarities between textual scenarios. (Mandelin et al, 2006) combines diagrammatic and syntactic heuristics for finding matches between architecture models. None of these were specifically designed for behavioural models and are either inapplicable or unsuitable for matching Statecharts models.…”

“…For example, they can be used interactively, with the developer seeding them with some of the more obvious matches, and pruning incorrect ones iteratively. Or, they can be customized for specific domains using learning-based techniques (Mandelin et al, 2006).…”