System Model-Based Definition of Modeling Language Semantics

In this paper, we present an approach to define the semantics for object-oriented modeling languages. One important property of this semantics is to support underspecified and incomplete models. To this end, semantics is given as predicates over elements of the semantic domain. This domain is called the system model which is a general declarative characterization of object systems. The system model is very detailed since it captures various relevant structural, behavioral, and interaction aspects. This allows us to re-use the system model as a domain for various kinds of object-oriented modeling languages. As a major consequence the integration of language semantics is straight-forward. The whole approach is supported by tools that do not constrain the semantics definition’s expressiveness and flexibility while making it machinecheckable.

💡 Research Summary

The paper introduces a novel approach for defining the semantics of object‑oriented modeling languages by grounding them in a comprehensive “system model.” Unlike traditional semantics that directly map language constructs to a fixed meaning domain or require fully specified models, this method treats the system model as a highly detailed declarative representation of object systems, encompassing structural elements (classes, attributes, inheritance), behavioral aspects (methods, state machines), and interaction mechanisms (communication channels, concurrency). Each language construct is expressed as a predicate over elements of the system model, turning the semantics into a set of logical constraints rather than a single, deterministic interpretation.

A key advantage of this predicate‑based formulation is its natural support for underspecified and incomplete models. Because the semantics are expressed as constraints, any element that lacks a concrete definition simply contributes a weaker predicate, allowing all implementations that satisfy the remaining constraints to be considered valid. This aligns with real‑world development where requirements are often vague and models evolve incrementally. For instance, a method may be declared without an implementation, or a communication protocol may be partially specified; the system model captures these gaps as open constraints, preserving flexibility while still enabling formal reasoning.

The authors also present a tool chain that makes the semantics machine‑checkable without limiting expressiveness. Semantic definitions are written in a domain‑specific language that allows unrestricted composition of predicates over the system model. The tool automatically translates these definitions into logical formulas that can be verified for consistency, completeness, and conflict detection. It further integrates with simulation and model‑checking environments, enabling validation that the defined semantics align with expected runtime behavior. This turns the semantic specification itself into a verifiable artifact, facilitating collaboration between language designers and domain experts.

Because the system model serves as a common semantic substrate, integrating multiple modeling languages becomes straightforward. Languages such as UML, SysML, and MARTE can each be mapped to the same underlying model elements, allowing cross‑language consistency checks, seamless model transformations, and unified analysis. The declarative nature of the system model also makes it compatible with existing formal methods, supporting automated code generation, test‑case derivation, and rigorous verification.

In summary, the paper contributes (1) a flexible, predicate‑based semantics that gracefully handles incomplete specifications, (2) a reusable, detailed system model that acts as a lingua franca for diverse object‑oriented modeling languages, and (3) a supporting tool chain that ensures the semantics are both expressive and machine‑verifiable. This work advances model‑based engineering by providing a robust foundation for semantic integration, evolution, and formal analysis across heterogeneous modeling ecosystems.