Abstract

UML models and OWL ontologies constitute modelling approaches with different strengths and weaknesses that make them appropriate for specifying different aspects of software systems. Since MOST will marry the two paradigms for exploiting their joint benefits, the core innovation aspect of WP1 is to extend and clarify the integrated conceptual picture and bring it alive with concrete specifications of metamodels, profiles, typical transformations and queries on the integrated models. WP1 will extend the existing metamodels and profiles for UML, OWL and related standards.

Mission

WP1 will provide the language layer for integrating models from the two paradigms. For this purpose, a joint metamodel will be provided. The possibility for querying over integrated models and methods for transforming instantiations of the joint metamodel into target plaforms will be provided by reusing and extending existing tools.

More specifically, MOST will integrate ontologies and software technologies at three different levels: metamodelling level, modelling level and tooling level.

1. Metamodelling level. In order to be able to query and transform integrated models, a unified view of such metamodels must be provided.
2. Modelling level. Ontologies and models will be integrated so that ontologies can be used in models, and vice versa.
3. Tooling level. Ontology based queries will be integrated on the level of MDA transformations. Queries will be provided to support the developer to fulfill requirements and address modelling decisions.

Use Scenarios

Figure 1 contains on the left side the world of ontologies and on the right side the world of MDSD models. In both worlds, several different metamodels are used (M2) as well as different metalanguages (level M3).

In the MDSD world, the major metalanguage is MOF (right side, level M3), so that the MDA model chain can be defined (right side, level M1), which connects the CIM (Computational Independent Model) via the PIM (Platform Independent Model) to the PSM (Platform Specific Model) and relies on consistently defined metamodels for all of them (right side, level M2).

Consider an example where some software, e.g. control application for network management, will first be designed as a CIM and some corresponding domain ontology may formalize allowed configurations in the target domain of networks. WP1 will provide a systematic connection between the CIM and the domain ontology, enabling a software module to ask for allowed configurations according to the domain ontology.

Context

Software engineering modelling languages and ontology languages use different approaches for describing abstractions of reality. Modelling languages tend to follow a layered view. An example is given by the threelayered approach of UML which differentiates the models, the superstructure and the metaobject facility (Figure 1, righthand side). This view fits perfectly to tool building where the models are the data, the metamodels act as data schemas and the metametamodel defines the overall approach.

In contrast, research in ontologies has not so much focused on a common syntax for different ontology languages, but rather on the exact definition of the semantics of these languages. As a first step to syntactic integration, the OMG has proposed the Ontology Definition Metamodel with, among others, MOF M2 metamodels for RDF and OWL. Nevertheless, ODM does not derive both languages from a common, integrated metamodel, which is a prerequisite for integrated model chains.

Using ontological concepts in UML models, deriving UML classes from OWL classes, or constraining UML classes with OWL expressions is only possible, if the languages are integrated much more tightly than proposed so far.

Also, querying integrated models, tailoring of reasoning (WP3), traceability (WP4), and process guidance (WP2) are only possible with an integrated approach. The gap which exists in this respect will be closed in WP1.

Investigation Problems

Languages for software models (like UML) and ontologies (like OWL) have different elements and properties and, hence, different metamodels. WP1 will allow for building and extending on the existing metamodels and UML profiles for OWL. Hence, WP1 aims at answering the following questions:

1. How can we integrate the UML metamodel and the OWL metamodel and UML profiles for OWL at the language layer?
2. How can we query the combined UML and OWL models and check for consistency in a practical manner at the metamodelling layer?
3. How can we use the combined UML and OWL models for transformation into models that yield effective ways of mixing imperative style programs (e.g. Java, C++, C#) with logical reasoning (e.g. queries to an OWL reasoner) abstracting the details away from the software developer?

Description of Work

1. Metamodelling. Propose a combined metamodel based on initial requirements from case studies.
2. Querying and Reasoning. Investigate the possibilities for querying the combined metamodel in order to access models in a flexible manner using existing languages.
3. Transformations. Explore the metamodelling of transformations that may be used in order to transform given combined model into more refined models (and eventually into code).