What are Ontolgies?

Ontologies are a concept borrowed from philosophy in order to facilitate the description of a field of discourse (Gruber, 1993). Corazzon (2003), in the introduction to his award winning site Descriptive and Formal Ontology, makes the distinction between ontology as conceptual analysis and ontology as technology. Within the context of this work we will be looking at ontology as technology rather than in the philosophical sense of the theory of objects and their ties, and it is probably wise to expand upon them in this context before progressing.

The realm of computer science was first introduced to the concept of ontologies in the framework of artificial intelligence. Researchers used ontologies to define the things which an AI could make judgments about, thus they provided an explicit specification of a conceptualization: the objects, concepts, and other entities that are assumed to exist in some area of interest and the relationships that hold among them (Genesereth & Nilsson, 1987). Wilson (1993) expounds at length about the inability of a model to completely describe an object and the same is true of an ontology; by necessity they must be a simplified description of the field of investigation. But by incorporating concepts from predicate logic they can be remarkably powerful tools to aid reasoning.

The author is conscious that to some extents the aim of this project is somewhat crippled in that the final product will be a tool which will aid the user as a human to have a greater understanding of, in this instance, an organisation whereas the generally stated use for the implementations of ontologies is to allow software agents to have an understanding of, and a tool for navigating, a field of discourse. This criticism is somewhat abated by the comprehension that an ontology will exist where before there was none. Indeed Noy and McGuinness (2000 p1) state that the reasons for developing an ontology are:

The first of Noy and McGuinness' points is perhaps the real reason for the development of any ontology, and this one in particular: Be it for the consumption of human or software agents, it will provide a means to navigate a common field of discourse by providing a lexicon understandable by all. Further, in conjunction with the aim of reuse of domain knowledge agents should be able to navigate outside of their sphere with some degree of assurance.

There seems to be something of a natural progression from these two ideas, i.e. the sharing of information and then the reuse of that information in other areas, providing a wider "sharing" if you will. This distribution of information is perhaps best illustrated by ontologies of time - where many systems would need to reference time as part of their field of discourse one single ontology can be used which the others could reference. This reuse of useful aspects of information is, at least in the mind of the author, something which is particularly close to the heart of computer scientists. Brooks (1995), in his seminal work The Mythical Man-month, talks at length of this concept of code reuse and present methods of describing ontologies are designed to make this process almost trivially easy to accomplish.

These two concepts are at the heart of Tim Berners-Lee's (2001) concept of the semantic web where he envisages a world where everyone's personal software agent is able to negotiate a maze of different fields and provide a solution to complex problems.

Making explicit domain assumptions serves to provide an avenue for altering those assumptions when domain knowledge is changed, either through an increase in knowledge about a domain or if the domain alters in some way, for instance should the management structure of the NHS Trust which is the focus of this work alter through some future act of legislation then the ontology would need to be able to change to reflect these alterations, this process must be as easy as possible to accomplish for someone without in-depth knowledge of computer coding. An explicit description also serves to allow new users a better understanding of the domain when they first become introduced to it.

The separation of domain knowledge from operational knowledge is also something which would be close to the ideals set out by Brooks. This means that an agent which is able to follow instructions associated with one ontology should be able to accept an ontology from a different field and make assumptions regarding that new ontology. Noy and McGuinness use an example borrowed from McGuinness and Wright (1998) and Rothenfluh et al (1996) of the development of a program which accepts the ontology of a PC manufacture and is able to reason through that ontology in order to create PCs being given an ontology of elevators and being able to produce elevators.

The analysis of domain knowledge is, in the authors own opinion, one of the primary reasons for the development of this particular ontology in that the ontology and its development serves as a paramount tool upon which to base reasoning in terms of the organisation of the organisation - if the reader will excuse the term. Even in the initial stages of the development of the ontology (of which more shall be said later) certain facts became obvious which might help to later prompt, analyze and structure re-organisation.

At its heart then an ontology isn't merely useful to develop for its own sake (despite the intellectual challenges it provides) but is a tool which can have many different purposes once completed, though whether an ontology can ever be termed as complete is debatable as it often serves as a window on a field of discourse at one particular moment of time only. It does however provide mechanisms whereby changes like those illustrated by Chalmers (1999), can be incorporated. Chalmers might be particularly interested in this inherent flexibility of ontologies when developing an ontology of scientific thought as it would be able to embrace those huge shifts of thought which he believes are a part of the development of any branch of science.

Ontologies are generally written using a language called XML and it is useful to examine this particular language in greater depth in order to better appreciate the context of ontologies in general and their place in the perceived development of information technology.