Ann Wrightson, February 2001
Topic Maps are an emerging technology - and like most emerging technologies, it is difficult at this stage to foresee which applications of Topic Maps will emerge as the "classics" of the technology. However, one thing is very clear: although most of the initial motivation, and the very early examples, came from publishing (encyclopaedias, conference proceedings, content-centred websites), the most valuable benefits will come from Topic Maps which are organizing corporate knowledge into more accessible, usable, integrated forms. This implies that these Topic Maps will be "seriously capable", i.e.:
So what does it take to have seriously capable Topic Maps? - there are several aspects, and this whitepaper gives you a brief tour around the essentials.
A Topic Map application's principal contribution within a distributed information system is to organize knowledge, information and data from one or more (probably quite different) sources, and do this in a way which satisfies some more or less well defined user needs.
This knowledge will typically have many different sources - the "base resources" of a Topic Map application. However, some of these resources will be pretty raw data; some will be more structured information; and some may be highly functional knowledge bases, which need integration with a wider repertoire of information.
Data -> Information -> Knowledge is a well-known progression, and there are technologies which already work pretty well within each area. However, a key benefit of Topic Maps is that by integrating disparate resources, Topic Maps can mediate this progression, and so enable the structured dissemination of targeted, usable knowledge throughout a diverse and demanding user community. In order for these demanding Topic Map applications to work, their structure must be carefully analysed and expertly designed. Existing design methods provide a place to start but as the technology matures, it is likely that distinctive design methodologies will be developed that are tailored to the effective deployment of these demanding and innovative knowledge technology solutions.
The resources to be organized by a "seriously capable" Topic Map application may be e.g.
Each of these will have its own characteristic structure. Many of these characteristic structures have their own inbuilt levels of abstraction (eg relational databases, various "server page" webservers) which support linking to individual resources and to derived information at various levels of granularity.
In addition, there may already be considerable investment in a particular company in ways of organizing information for specific purposes, for example business process models; intranet directories and site maps; customized searches and filters. A Topic Map application must be able to make use of these, providing additional capability rather than throwing everything back to the drawing board.
Fortunately, the open architecture provided by Topic Maps is very flexible in this regard. For example, if you have a customized interface to a bought-in information service, then you can integrate it at that interface, and retain your investment. Perhaps more acutely, if you have a pressing need to integrate your proprietary knowledge management system with the intranet of a newly acquired sister company, in order to realize the synergies justifying the business acquisition - then a Topic Map application can not only interrelate the two, but also make both available as resources for future "webspeed" developments.
The overall architecture of this kind of Topic Map application is a collection of resource bases, with each resource base clothed in a "Virtual Topic Map" - an interface which makes that particular resource base look like a Topic Map with a suitable structure.
In order to construct such a "Virtual Topic Map", there must be some means of identifying the topics, the topic occurrences for each topic, and other aspects such as scoping. In some cases, these Topic Maps will be crafted by hand - but in larger applications, there must be some way of computing the Topic Map (eg by algorithmic searches based on keywords derived from a designed ontology).
So in a nutshell, the first step in bringing a resource-base into a Topic Map application is to determine how it needs to be "visible", and then build a Topic Map adapter layer over it embodying that view. Where the purpose of Topic Map application is topical organization of the resources, then this may be all there is to do. (However, even in this case I'd expect some other linkages, eg to a thesaurus or subject index.)
The second step is to provide a further Topic Map layer interrelating the different structures emerging from the various resource bases, and integrating them into whole which achieves the overall aim. This is where the real value of a seriously capable Topic Map application will be delivered. The structures used will depend on the nature of the business activity to which the Topic Map application is expected to deliver value, eg corporate knowledge management; a consumer web-portal; or supply chain integration. It is at this level, too, that Topic Maps will be able to integrate with more specialized AI-based Knowledge Representation and inference tools, such as Case-based reasoning and Expert systems.
The nature of the topics in the second-level Topic Map, and their associations with with topics which provide a view of the resource base(s), must be both well designed, and capable of being automatically generated. It's a bit more likely that this second layer will be built by hand - or based on content structures and metadata which are generated by existing well-designed and implemented business processes, such as editorial control of electronic content, or integrated sales, marketing and service delivery.
Topic Map system integrators - or rather, knowledge-integrators - will, as they gain experience, develop a toolkit of reusable structures to use at this level, effectively providing and supporting innovative kinds of knowledge management for the growing community of Webspeed businesses.
Copyright Ontopia AS, 2000