Semantic Knowledge Representations for Soft Data Fusion - Efficient Decision Support Systems: Practice and Challenges from Current to Future - page 10

can be distinguished in a semantic network: concepts (or classes of objects) and individuals
(or instances of the classes). The relations represented by the edges of the network are, among
others, instanciation (“is a kind of”) and composition (“has a”).
The semantic networks were a first attempt to formalize semantic information and the
reasoning process. However, as automatic processes attempted to get smarter, new semantic
representations with more expressiveness and formalization were developed.
4.2 Ontologies
Ontology was initially a field of metaphysics, which aim is to study the kinds of things that
exist and the relationships that can be observed among these things. Applied to the field of
computer science, an ontology is a structured set of concepts that model a specific domain.
Since the mid 70's, AI research scientists have had the need to capture knowledge, in order to
provide “intelligent” capabilities to computers. Studies were achieved that aim at providing
the ability to store general and domain specific knowledge, in a way that is understandable
both by humans and computers.
An ontology captures the model of a specific domain as a catalog of categories of entities
and the semantics associated with this domain. This allows making inferences and reasoning
about the domain. The main components of an ontology are:
• individual entities,
• classes, which are sets or collections of entities
• attributes, which are properties of the entities represented by the different classes
• relations, which are relationships among different classes and
• events that change the state of some of the properties of entities or that modify
relationships among entities.
Within the fusion community, works such as Matheus et al. (2003) insist on the importance
of using ontologies to represent knowledge. For the military domain, several ontologies were
developed, such as the JC3IEDM2 ontology (MIP (2005)) and the “SAW Ontology” described
in Matheus et al. (2003).
4.3 Conceptual graphs
The conceptual graphs formalism is a model that encompasses a basic ontology (called
vocabulary ), graphs structures and operations on the graphs. The vocabulary defines the
different types of concepts and relations that exist in the modeled application domain, while
the graphs allow representing observations.
The conceptual graphs where proposed by John Sowa in Sowa (1976) as a graphical
representation of logic, which was inspired by Peirce Peirce (1932). They allow representing
knowledge in a easily readable manner for humans, experts of specific application domain,
but non experts of knowledge representation formalisms. In our work, we use the conceptual
graphs. The numerous studies achieved regarding graph algorithms and conceptual graphs
in particular (Chein & Mugnier (2008)), lead us to use this model.
5. Using semantic representations for soft data fusion
5.1 Case study
We applied our proposition for soft data fusion within a TV program recommendation system.
While the number of channels that one can access increases very fast, the aim of the system