4.15 Technologies and explicit knowledge continued
We noted earlier that, in philosophy, an ontology refers fundamentally to ‘being’, or ‘what can be’. In the field of artificial intelligence the term ‘ontology’ has been appropriated to mean a ‘reusable terminological scheme’ or, if you prefer, a ‘conceptualisation’: a scheme for providing a rigorous description of the concepts, attributes and interrelationships deemed relevant to describe a particular aspect of the world. Its precision means that it can serve as an agreed model of a domain to ensure a common point of reference between parties. An ontology is an abstract knowledge model in the sense that, like an agreed standard, it does not need software in order to exist (it is a specification of how to talk about the world). However, a strength is that, if the ontology is created in digital form, software tools can assist in checking its internal consistency, and can convert it into a knowledge-based system for a particular application to a problem.
Consider the example shown in Figure 10, taken from a healthcare support system.
If someone has already invested much effort in deliberating about the ontological structure of a particular domain, others who are seeking an understanding of that domain could benefit from this, rather than having to reinvent the wheel. Numerous knowledge-sharing initiatives are under way, developing notations for expressing ontologies, which are then published and exchanged over the Web. A common notation makes it possible to exchange ontological structures that can then be embedded into someone's own ontology, if they are using the same notation. (In reality, variations in the use of language and approach usually entail some tailoring to make someone else's ontology ‘plug into’ your own.)
Part of the notational specification (using the Ontolingua notation: Gruber, 1993) for the graphical ontology structure in Figure 19 is shown in Box 4.14. It is immediately obvious that only specialists will wish to read and write such specifications. The notation is formal to enable automatic generation and interpretation of ontologies by computers, but at the same time knowledge engineers can read and write such code. The services (such as ‘smart email’) enabled by having a partial model of the world would then be delivered to end-users using conventional user interfaces.
Box 4.14 Part of the specification for the graphical ontology structure in Figure 10
In principle, we would expect notations and tools for modelling the structure of specialist knowledge to have great potential for knowledge management. At present, however, knowledge modelling is still largely in the research laboratories. The effort and expertise required has meant that few ontologically based technologies have been widely used on a significant scale. A critical factor that will determine whether ontologically ‘enriched’ technologies are adopted is the usability of the system. How much understanding of the ontology is required to use it effectively? Can new entries be added easily to the knowledge base at the right moment? We may see technology vendors beginning to embed more explicit representations of knowledge in their systems in the next few years, particularly given the high profile that two related technological developments are receiving: metadata, as discussed above, and software agents, which are discussed later in this section.