Abstract
As data exchange and model transformation become ubiquitous nowadays, it is a key requirement to
improve interoperability of enterprise systems at the semantic level. Many approaches in
Model-driven Architecture (MDA) and Model-driven Interoperability (MDI) emerge to fulfil the above
requirement. However, most of them still demand significant user inputs and provide a low degree of
automation,
especially when it comes to finding the mappings. A generic approach that can easily handle both
semantic interoperability and automatic transformation is currently missing.
This paper presents AutoMapping, a semi-automatic model transformation architecture. This approach
focuses on two aspects: 1) semi-automatic mapping between data models expressed as class diagrams
by involving minimal user interactions at design-time; 2) generation of executable mappings.
Particularly at design-time, a semantic engine that solves various kinds of semantic attribute
mismatches is devised, such as type, scale, synonym, homonym, granularity, etc. Furthermore, a
heuristic-based similarity analysis between each pair of classes is proposed, which takes all
relations of classes into account, such as inheritance, reference, etc. Finally, a method is given
to match fragments and then generate mappings specification that conforms the proposed mapping
metamodel for solving existing semantic mismatches.
The main contribution of this paper is to create a generic platform-independent approach for semi-
automatic model transformation towards semantic interoperability, with tool-based implementation
and motivating case experiment, showing the feasibility of using MDA and MDI techniques for
semantic
improve interoperability of enterprise systems at the semantic level. Many approaches in
Model-driven Architecture (MDA) and Model-driven Interoperability (MDI) emerge to fulfil the above
requirement. However, most of them still demand significant user inputs and provide a low degree of
automation,
especially when it comes to finding the mappings. A generic approach that can easily handle both
semantic interoperability and automatic transformation is currently missing.
This paper presents AutoMapping, a semi-automatic model transformation architecture. This approach
focuses on two aspects: 1) semi-automatic mapping between data models expressed as class diagrams
by involving minimal user interactions at design-time; 2) generation of executable mappings.
Particularly at design-time, a semantic engine that solves various kinds of semantic attribute
mismatches is devised, such as type, scale, synonym, homonym, granularity, etc. Furthermore, a
heuristic-based similarity analysis between each pair of classes is proposed, which takes all
relations of classes into account, such as inheritance, reference, etc. Finally, a method is given
to match fragments and then generate mappings specification that conforms the proposed mapping
metamodel for solving existing semantic mismatches.
The main contribution of this paper is to create a generic platform-independent approach for semi-
automatic model transformation towards semantic interoperability, with tool-based implementation
and motivating case experiment, showing the feasibility of using MDA and MDI techniques for
semantic
