Technical Report 303, c4e-Preprint Series, Cambridge

Authors: Arkadiusz Chadzynski, Heidi Silvennoinen, Ayda Grišiūtė, Feroz Farazi, Sebastian Mosbach, Martin Raubal, Pieter Herthogs, and Markus Kraft*

Reference: Technical Report 303, c4e-Preprint Series, Cambridge, 2023

Associated Themes:
 

• Revised novel semantic web systems architecture for inferences.
• Inference Agent capable of deriving new knowledge by applying graph and ontology based inference algorithms.
• OntoInfer ontology that encodes taxonomy of existing knowledge graph reasoning algorithms linked to the agent's tasks.

This paper presents revised novel semantic web systems reference architecture for inferences and components that can store and operate on knowledge in the form of a fully dynamic graph to infer new statements by an intelligent autonomous agent capable of making informed choices based on long-term memories about its tasks that implement inference algorithms of all currently known classes. An Owlconverter tool was designed and developed as a new component which can produce fully dynamic knowledge graphs without information loss that otherwise occurs while attempting to store complex concept definitions in existing open-source dynamic RDF stores. An Inference Agent with extended cognitive capabilities of making informed choices based on long-term knowledge was designed and developed to act as an extended inference engine supporting all currently known classes of knowledge graphs inference algorithms. This capability is supported by the newly developed OntoInfer ontology that encodes the taxonomy of those algorithms linked to instances of the agent's tasks allowing the agent to make choices based on the knowledge stored in the knowledge graph. This extended architecture can demonstrate the implementation of tasks designed to work as independently executed threads containing examples of known inference algorithms using existing libraries and reasoning engines (Jena Jung and HermiT). Multi-domain reasoning capabilities on city object descriptions in terms of OntoCityGML, OntoZoning and OntoBuildableSpace were showcased on plot data provided by Urban Redevelopment Authority of Singapore converted into OWL 2 compliant knowledge base.

PDF (2.8 MB)