This thesis provides a detailed description of the research undertaken into the creation of a framework that uses Semantic Web languages to implement a recently developed commonsense reasoning formalism called Discrete Event Calculus (DEC). It aims to show to what extent DEC reasoning can be applied to Semantic Web data, using the Semantic Web standards and supporting development environments available for the purpose in 2008, when the research programme commenced.The research aims to provide an accurate and reusable DEC ontology using the languages defined in Semantic Web Standards. To this end, an ontology describing the DEC entities and axioms is defined in OWL and SWRL; this represents the core elements of the DEC formalism, namely its set of logical types and predicates and the relations between them. The ontology is used together with a proof-of-concept DEC resolver software that applies the ontology to an existing rules engine, so that new inferences can be created from a DEC domain. The design and implementation of the combined ontology and software framework are described in detail. The methodological issues involved in reconciling a software model with an ontology model are also discussed and the capabilities of the framework are validated by a series of tests modelled on established AI benchmark scenarios that can be resolved correctly using DEC. The results confirm that the framework will create the appropriate inferences with reference to the benchmark problems, though they also highlight some of current limitations in the framework, notably to do with how it represents changing fluent values.A detailed sample domain ontology is provided, which is based on the domain of turn-based multiplayer online games; this illustrates how the DEC ontology defined in this research could be extended for use with other domains. A further extension of the DEC ontology is proposed, which enables the resolver to represent real-world time values independently of the timepoints defined as part of the formalism.Finally, the strengths and extant boundaries of the chosen approach are discussed and suggestions are provided for improvements that could form the basis of future work.
|Date of Award||2010|