The Knowledge Management Connection   

Every organization, whether centralized or distributed, that must deal with large amounts of information inevitably feels compelled to create a representation of the concepts of that organization. At the moment, most organizations that do so are rather large —  for example, the medical community and the Medical [Area] Subject Headings (MeSH) or the engineering community and its EJC Thesaurus of Engineering and Technical Terms. However, we need representations of concepts even in small business organizations, for the following reasons:

• You cannot take effective action with predictable results unless you know the meaning of the ideas in the organization with some degree of precision. Ineffectiveness of activities in knowledge-based organizations is often traceable to absence of a shared understanding of the domain. The need for precision in most business activities may be less than in the hard sciences or medicine, but it is far greater than in much humanities research, in which precision of meaning is considered essential.
• You cannot perform effective, predictable operations on those concepts with computers (for example, identifying duplication of information) — in order to leverage their value — without making them explicit and modeling them precisely.
• Evaluations of the value of concepts and content are truly important contributions to the knowledge. You have to make it a priority in your modeling and implementation!!!!
• Clustering documents together by semantic similarity by itself fails to meet the requirements of effective access to, and transfer of, new knowledge. Documents require too much effort to read and interpret, and they can be interpreted badly. You cannot transfer knowledge effectively unless you make it explicit and precise. This is especially true in an environment characterized by cross-functional teams, cross-domain areas of inquiry and partnering, rapid change, and distributed collaborative work. Does that sound a lot like today's business environment???
• Representing concepts explicitly unmasks hidden assumptions. Part of this benefit is achieved by abstraction and creation of “structural knowledge” — by separating concepts and the persistent semantic relationships among them from the many specific instances in which they occur.
• Building new knowledge rests on what has gone before. “The notion of incrementally building knowledge on previous knowledge presupposes a continuing strain of topicality that binds that knowledge together. While this view of knowledge growth is more often associated with the natural sciences, there is no field in which scholarship can be ignorant of its predecessors.” (Rebecca Green, “The role of relational structures in indexing for the humanities.” Knowledge Organization, 24 (1997) No. 2. Pp. 72-83)

  Organizational knowledge is cumulative. As the amount of raw information expands, much of the detail becomes irrelevant and the need for abstraction — the ability to see the essentials, what is important, and what is persistently true — increases dramatically. A representation of the concepts of a domain meets this requirement.

• Representing organizational knowledge as a set of concepts and relationships among concepts addresses a common requirement of information seekers: they have an “anomalous state of knowledge.” (N.J. Belkin, “Anomalous states of knowledge as a basis for information retrieval.” Canadian Journal of Information Science, 5, 133-143.) That is, you often don't know precisely what you want to know. Chances are, what you actually want is related to what you ask for. And what you are looking for is instances of related ideas, not related documents. (It would be virtually impossible to reach any kind of consensus about the relationships among documents, because each document contains many ideas, only some of which are relevant to the reader at any particular time.)
• Organization of concepts should be independent of language. Even within one language, a system of knowledge representation should permit any number of names for one concept. The concept remains unique, no matter what name is used — just as a person may have several names, but only one identity. (Conversely, one word may have many meanings — some slightly different, others completely disconnected.)
• Relationships among concepts are/should be persistent. Jobs and processes change, but much knowledge does not.
• Superficial characteristics don't tell you  about the essence of things. For example, there is a species of parrot in which the male is a bright red and the female is a bright green —  not the usual major/minor variation in color between the sexes of birds, but an entirely different bright color. This apparently confused ornithologists for some time … but not the parrots, who knew the essential difference.

There is only one proven way of making meaning precise — abstracting the concepts from a body of knowledge and organizing them in relationship to each other in a formal structure.

Ultimately, groups and individual map their own knowledge representations over a body of documents. This is entirely natural. Medical doctors, biochemists, and biologists, for example, will select different concepts — and different relationships among those concepts — for their preferred representations of a bio-science library.