Transcript Document
Semantics-Based Aspect Composition
Participants
• Awais Rashid, Lancaster University, UK • Ruzanna Chichyan, Lancaster University, UK • Ana Moreira, New University of Lisbon, Portugal • Mansour Zand, University of Nebraska Omaha, USA • Jiang Ningkang, E China Normal University, Shanhai • Wu Yan, University of Nebraska Omaha, USA • Jon Whittle, George Mason University, USA
What is Semantics-Based Aspect Composition?
1. Leveraging the semantics of the underlying language in composition 2. Defining a semantics for composition languages 3. Composition using the semantics of the domain of the application We focused on (3)
Approaches Discussed
• Lancaster approach for AORE – Natural language semantics (e.g., verb categories, synonym lists) • Whittle/Araujo/Moreira – Classification of concepts • Domain-specific modeling languages (DSMLs) – Comes with semantics of concepts but may be limited to specific domains
Where does the semantics come from?
• Semantic information usually not present – Must be explicitly given or derived • Explicitly given: – Annotations/DSMLs/etc.
• Derived: – Alpha (TU Darmstadt) – From natural language semantics
Advantages
• Decouples the aspect: don’t refer to name or number that may change • Could potentially avoid many of the consistency problems associated with composition
Avenues for Future Research
• Trade off of providing the semantic information versus the ROI in using it – Cf.static analyzers • Relevancy: given an ontology, how can I quantify how relevant or related concepts are to each other? Chair->session chair->table. Could possibly use context?
• Claim: semantics-based architecture composition less well understood than requirements • DSMLs: – Semantics usually implicit in code generation tools – Domain-specific composition strategies