Perhaps the most fundamental notion for successful large-scale collaboratively-built systems is Semantic Interoperability. This talk defends the view that to meet the requirements for full semantic interoperability, we need to create an engineering discipline that is based on the followings tenets:

• (i) real-world semantics should take precedence over pure formal semantics;
• (ii) real-world semantics is very much about “Ontology” in the true original sense of the term;
• (iii) before considering automated reasoning and computational non-functional requirements (e.g., computational tractability), semantic representation systems must strive for precision, expressivity, truthfulness to the underlying conceptualisation, and pragmatic efficiency;
• (iv) we need proper foundational theories and engineering (methodological, computational) tools based on these.

The talk presents the results of a two-decades research program aimed at addressing these issues. The program is based on two fundamental pillars: the development and use of foundational axiomatic theories (foundational ontologies) and principles in the design of semantic models (domain ontologies, knowledge graphs), and the use of patterns as central complexity management tools in this process. After introducing an Ontology Pattern Language (OPL) for Semantic Modelling that is truly ontology-based, I will discuss Ontological Anti-Patterns (OAPs) as fundamental tools for model validation. Moreover, I will elaborate on the role of having execution semantics of models for model understanding, and validation, and how this approach associated with a particular machine learning technique can support sustainable Anti-Pattern detection and rectification as well as model evolution.