Understanding Semantics and Ontologies: They're Quite Simple Really - If You Know What I Mean!

Over the past few years, it seems that a good way to ensure a small audience for a technical session at any of the mainstream GIS conferences, is to include in your presentation title words such as `semantics', `ontologies' or `interoperability'. While these topics continue to be popular at GIScience research conferences and in the scientific literature, there is little doubt the average GIS user finds them both confusing and uninteresting. Yet closer examination reveals they are in fact quite relevant to GIS users in the 21st century. Clearly, then, the problem lies in the message delivery ± not its content. Accordingly, in this editorial I would like to help demystify these subjects and hopefully explain them in more practical and simple terms. To begin with, let's go back in time to the beginnings of GIS. If you were trying to put together a dataset for use in a GIS about 25 years ago, the likelihood was that it was for a public agency as there were few private sector systems at that time. Furthermore, in those days data modeling was relatively straightforward. If you worked in the natural resource arena you chose a grid-cell approach, and if you worked in utilities you opted for the vector model. As you set about the major task of digitizing your hardcopy data, you were probably too busy to be concerned with what other agencies were doing with their GIS, or how they were going about the task of representing and categorizing the real-world within their databases. For example, in your database you only defined two classes of roads because that was all you were interested in and you represented them by line segments. However, the local transportation agency had decided to define five classes (because it needed them for capital funding and maintenance program purposes) and to represent them by polygons, while the state land registry was simply digitizing land parcel boundaries ± and it just happened that the regularly spaced gaps left between the street blocks comprised their interpretation of roads (identified simply by placing text annotation in the database). So we have three different examples of how roads might be modeled and classified, each of them quite valid and sensible within the context of their host agencies, and no doubt similar stories can be told in most jurisdictions around the world. By and large, these inter-agency differences went unresolved throughout the 1970s and early-1980s, as lengthy and expensive project implementation plans crept toward finalization. In any case, data sharing was not common back then because the telecommunication networks and transfer protocols had not been widely established ± and remember, this was the era when data transfer meant physically carrying delicate Transactions in GIS, 2002, 6(2): 83±87