DIS: A Data-centred Knowledge Representation Formalism

The Domain Information System (DIS) is a novel data-centered knowledge representation formalism that offers a modular structure, which separates the domain knowledge (i.e., the ontology) from the domain data view. The formalism is designed to take advantage of a Cartesian perspective on information, using part Of as its fundamental ontological relation. A DIS consists of a domain ontology, to model the domain knowledge, a cylindric algebra, to model the domain data view, and an operator which couples the two. The core component of the ontology is a Boolean lattice built from atomic concepts taken from a data schema. It is enriched with relevant concepts from the domain of application. A detailed case study is presented to highlight the features of this formalism. The integration of several datasets and their respective ontologies for reasoning tasks requiring data-grounded and domain-related answers to user queries is illustrated. The ability to handle information evolution is demonstrated. Two open issues in ontologies, namely, the lack of clear and explicit guidelines for ontology construction, and the prohibitive cost of adapting and reusing existing ontologies, are addressed.