Knowledge Representation in the Semantic Web

Clearly the basis of the semantic web is a collection of information, rules and representations that can be used to conduct automated reasoning; a task that lends itself to the techniques used in knowledge representation.

Knowledge representation defines how we might represent the knowledge that is stored in our brains. We can then use inference rules and systems to extrapolate new knowledge from this. This forms a system of defining objects and the relationships between them, known as an ontology.

However, ontologies are at their most useful when they are considered as a group. Ontologies must be linked together, even derived from each other, and should be able to change over time in response to trends.

The problem lies in the fact that traditional knowledge representation systems and structures worked on the basis of all users having exactly the same definition of everything. Each knowledge representation system had its own rules for making inferences about its data. This limited the scope of these systems, as data and rules could not be transferred from one system to another, since they existed in entirely different forms.

In contrast, the semantic web must use a far more versatile language. Rules and definitions must be allowed to have multiple meanings (just like the English language!), be defined in multiple ways, and rules must exist in multiple forms. The more expressive the language for the rules, the more widely and resourcefully we can reason. These rules must not simply be new rules, but must be importable from all existing knowledge-representation systems.

An ontology 'provides an explicit conceptualization that describes the semantics of the data' (Fensel, 2004). Therefore the ontology is a specification of the rules and conventions that a network of information will use to represent itself.

An ontology usually specifies things that are relevant to the program in question, for example things it must consider as 'facts' such as: