T2B2T: The Ontology for Adaptive Agent-Driven Seamless Integration with the Semantic Web

Open science research, enabled primarily by shared Knowledge Graphs (KGs) whose backbone is the so-called Semantic Web, offers great opportunities to infer new information from prior-gathered data. Through the SPARQL language, it is possible to filter and process extracted data, from which implicit triples can be inferred, using reasoners such as Hermit, Pellet, etc. and also SWRL axioms. However, some KGs may lack the triples necessary for a specific research task and calculating these triples may require functions beyond SPARQL and OWL 2-based reasoners, which forces agents designers to delegate such task to high-level programming languages. In case of investigations by agent-based modeling involving the Semantic Web, these new triples must be integrated into KGs accordingly, otherwise a not seamless integration may invalidate aggregated outcomes given by interrelated inferences, in either mono- or multi-agent setup. In this paper we introduce the novel paradigm Triples-to-Beliefs-to-Triples (T2B2T) to model agent-based seamless integration with the Semantic Web, in order to adapt KGs and enable them to support Belief-Desire-Intention (BDI) inferences. Afterward, beliefs can be newly translated into triples, which will populate the derived KGs endowed with a legacy of past task-oriented inferences, achieved in the domain of BDI-logic.