Concept-based Management of Technical Knowledge

Knowledge about technical systems deserves our attention for several reasons. It shapes industry’s core processes which depend on knowledge. It is an asset of great value because it is used by those who engineer, operate or maintain a system. This paper explains key characteristics of technical knowledge, the reason why it is difficult to manage and what we can do about it. Our experiences e.g. with documentation of complex communication systems and extensive use of SGML let us believe that management of technical information should rely more on conceptual structure and semantic markup. This will affect the way we design DTDs and use ontologies. Characteristics of Technical Information A prominent feature of technical information is it‘s complexity and we are greatly challenged to reduce it. An example will illustrate this case: the documentation used by technicians to operate and maintain a communication network. We generated such documentation as a technical information base of 20000 HTML-files and 100000 links among them. The major part of it describes service tasks and procedures; the rest describes functions, components, interfaces, dependencies, versions and variants of network elements. This information is carefully managed throughout its lifecycle. The authoring process, for instance, is supported by tools which prescribe how to structure information. To enforce structure we use five SGML-DTDs each of which consists of about 300 elementand attribute-types. All descriptions are stored as SGML-instances in databases; generators select and assemble the information pertinent to a specific version of a product and produce various publication forms. One such publication form is HTML as mentioned before. So, it seems obvious that the complexity of technical knowledge stems from it‘s volume and structure. There is however still another dimension of complexity which is of a conceptual nature. To explain the point we will sketch a piece of information flow. Before product information (e.g. Copyright © 1999, American Association for Artificial Intelligence (www.aaai.org). All rights reserved. the description how to operate a network) arrives at the customer’s site, it flows through several departments of the manufacturer of network gear. Typically, the information flow from engineering to service departments „transports“ numerous aspects of a product: how can the product be installed, configured or diagnosed; how can the product interface to those foreign systems which constitute the customer environment; other views concern the states or dynamics of hardware or software. Last not least some views concern a product's legal and economic properties. The views and concepts which structure technical knowledge constitute the „conceptual interface“ when information flows from one department into another. Such views and concepts must be known by the receiver of information to understand the sender. In our domain we deal with an estimated 10.000 significant concepts and many more technical terms. Communications is like most other fields of technology. It is rich in terms and concepts which, however, have never been systematically collected. But terms and concepts could be put to use immediately, for instance to build a portal to access the information on our family of products. Vertical portals may become part of the web technology catering to the broadening clientele of technical information.