DOGMA, short for Developing Ontology-Grounded Methods and Applications, is the name of research project in progress at Vrije Universiteit Brussel's STARLab, Semantics Technology and Applications Research Laboratory. It is an internally funded project, concerned with the more general aspects of extracting, storing, representing and browsing information.[1]
Contents |
DOGMA [2] is an ontology approach and framework that is not restricted to a particular representation language. This approach has some distinguishing characteristics that make it different from traditional ontology approaches such as (i) its groundings in the linguistic representations of knowledge[3] and (ii) the methodological separation of the domain-versus-application conceptualization, which is called the ontology double articulation principle[4]. The idea is to enhance the potential for re-use and design scalability. Conceptualisations are materialised in terms of lexons. A lexon is a 5-tuple declaring either (in some context G):
Lexons could be approximately considered as a combination of an RDF/OWL triple and its inverse, or as a conceptual graph style relation (Sowa, 1984). The next section elaborates more on the notions of context.
Another distinguishing characteristic of DOGMA is the explicit duality (orthogonal to double articulation) in interpretation between the language level and conceptual level. The goal of this separation is primarily to disambiguate the lexical representation of terms in a lexon (on the language level) into concept definitions (on the conceptual level), which are word senses taken from lexical resources such as WordNet[5]. The meaning of the terms in a lexon is dependent on the context of elicitation [6].
For example, consider a term “capital”. If this term was elicited from a typewriter manual, it has a different meaning (read: concept definition) than when elicited from a book on marketing. The intuition that a context provides here is: a context is an abstract identifier that refers to implicit and tacit assumptions in a domain, and that maps a term to its intended meaning (i.e. concept identifier) within these assumptions.[7]
Ontologies naturally co-evolve with their communities of use. Therefore, in De Leenheer (2007)[8] he identified a set of primitive operators for changing ontologies. We make sure these change primitives are conditional, which means that their applicability depends on pre- and post-conditions[9]. Doing so, we guarantee that only valid structures can be built.
De Leenheer and de Moor (2005) distinguished four key characteristics of context:
Based on this, they identified three different types of context dependencies within one ontology (intra-ontological) and between different ontologies (inter-ontological): articulation, application, and specialisation. One particular example in the sense of conceptual graph theory[10] would be a specialisation dependency for which the dependency constraint is equivalent to the conditions for CG-specialisation[11]
Context dependencies provide a better understanding of the whereabouts of knowledge elements and their inter-dependencies, and consequently make negotiation and application less vulnerable to ambiguity, hence more practical.