Easing Model Transformation Learning with Automatically Aligned Examples

Model Based Transformation Example (MTBE) is a recent track of research aiming at learning a transformation from examples. In most MTBE processes, a transformation example is given in the form of a source model, a transformed model and links between source elements and the corresponding transformed elements. Building the links is done manually, which is a tedious task, while in many cases, they can be deduced from the examination of the source and transformed models, by using relevant attributes, like names or identifiers. We exploit this characteristic by proposing a semi-automatic matching operation, suitable for discovering matches between the source model and the transformed model. Our technique is inspired by and extends the Anchor-Prompt approach, and is based on the automatic discovery of pairs of anchors (pairs of elements for which there is a strong assumption of matching) to support the whole matching discovery. An implementation of the approach is provided for validation on a case study.