| Technologies, techniques,
and disciplines in knowledge management
Part 5 of "New language for new leverage: the terminology of knowledge management" (Philip C. Murray) / Torna a Indice Strumenti |
Many of the concepts of "knowledge management" reach much further back in technical disciplines than their recent "discovery" by business management consultants. The term knowledge management -- in its current generally accepted meaning -- has been with us for at least 13 years, and perhaps much longer.
The computer-supported collaborative work (CSCW) community has been addressing issues of shared development of knowledge artifacts for many years. The community has at least one annual conference in the United States. Groupware is sometimes used as a synonym for CSCW, and Lotus Notes often appears to be the defining CSCW application -- even though there are other groupware products, including Netscape's Collabra Share.
The term artifact-based collaboration is often used in association with CSCW technology because the result of the activity is an artifact -- for example, a document authored by many people.
Recent developments in corporate intranets (q.v.) are likely to dramatically increasing the level of interest in CSCW, as IP-based technologies replace or complement proprietary products like Notes. See Hal Richman's "Why the Lotus Notes vs. intranets debate shouldn't be about technology" in this issue of KM Metazine.
Distributed hypertext systems -- and, in particular, open hypertext systems -- have been concerned with the generation and leveraging of organizational knowledge for more than a dozen years.
Theodor Holm Nelson coined the term "hypertext" in the 1960s, and his writings about representation, access, and management of knowledge -- embodied in his vision for Project Xanadu, a global "docuverse" that pre-figured the World Wide Web -- still deserve close reading (no matter what you may have read in Wired about Nelson).
Hypertext pioneer Douglas Engelbart's Augment hypertext system helped "augment human intelligence" at McDonnell-Douglas Corporation by supporting on-line sharing of knowledge resources.
In 1983, hypertext pioneers Rob Akscyn and Don McCracken, aware of Peter Drucker's term knowledge worker, named their landmark distributed open hypertext system the Knowledge Management System (KMS). They wanted to convey to their new commercial audience that their product was designed to manage the knowledge assets of business organizations. (Prior to 1983, KMS had been called "ZOG" and was actually put to use on a U.S. Navy aircraft carrier.) Both "Knowledge Management System" and the abbreviation "KMS" are trademarks of Knowledge Systems Inc.
The World Wide Web is a distributed hypertext system, but it is not "open" because Web pages are "read-only." Although WWW creator Tim Berners-Lee originally conceived the Web as an open system for research, you cannot modify the content of someone else's Web page or even attach a comment directly to the page.
Document management systems originally were primarily concerned with providing online access to documents stored as bit-mapped images. The fastest area of growth in the past few years, however, has been in the area of "compound document management" technology, which addresses managing document content at the level of smaller components.
Document management technology -- already in widespread use in large, information-intensive companies-- is likely to become an integral part of virtually every "intranet" in one form or another.
See "`Thin' Mozilla overwhelms `fat' management technology; film at 11:00" and "Information, knowledge, and document management technology" in preceding issues of KM Metazine.
Geographic information systems, a term associated with knowledge management, is used with what appears to be two quite different meanings. A recent CIO article uses it in the sense of a graphic tool for knowledge mapping:
Known as geographic information systems -- GIS for short -- the technology involves a digitized map, a powerful computer and software that permits the superimposition and manipulation of various kinds of demographic and corporate data on the map.
- Srikumar S. Rao, "Corporate Treasure Maps: The days of sticking colored pins in wall maps are numbered." From: A Special Report Supplement: Information Technology and the Bottom Line. The FW/CIO Quarterly, p. S1, CIO, July 1995.
This usage of geographic information systems seems to conflict with its original usage in association with metadata (q.v.) . The graphic nature of maps made them a natural candidate for retrieval by key words and other formalisms.
Help desk technology is primarily concerned with routing requests for help from information seeker to the right technical resolution person within an organization. The help desk may be internal or serve a customer-support function. Although this relatively new category of products is often most closely associated with corporate networking, interaction with internal and external customers can provide a key source of knowledge about desirable characteristics of future products as well as improvements for current products.
The help desk call logs are a gold mine if abstracted, analyzed, organized and made available. The trick is to use the same knowledge to solve problems, train and assist with decisions. This requires greater intelligence at the interface than currently available.
People who outsource their help desks are giving away hidden opportunities to learn even if the sub-contractor must report back. The value lies in the unobtrusive relationships and the chance to cement ties with internal and external clients. It requires real knowledge of the business and the culture to read the help desk signs correctly and to mine the call logs efficiently.
- Source: email from Denham Grey of GreyMatter, Inc.
Intranets -- intra-corporation networks that use the Internet's IP (Internet Protocol) standard -- not only permit sharing of information, but they also view the organization's information (including structured resources like relational databases as well as unstructured text) through Web browsers like Netscape Navigator.
The use of a hypertext browser as the entry point for computing -- widely discussed in conjunction with "Internet boxes" -- is an interesting inversion of the traditional computer interface model, in which computer users run applications like word processors and relational databases in order to view and process information.
Once again, though, the idea is not new. While furthering the development and commercialization of the Knowledge Management System at Scribe Systems, Rob Akscyn proposed KMS as a general interface model in a formal response to an Open Software Foundations System Request for Technology. (The OSF ultimately chose Motif.)
Knowledge representation -- explicit specification of "knowledge objects" and relationships among those objects -- takes many forms, with variations in emphasis and major variations in formalisms.
Knowledge representation allows computers to reconfigure and reuse information that they store in ways not narrowly prespecified in advance.
- From "Building Global Knowledge Webs; Knowledge Representation for the Web." Panel Session at the Fourth International Conference on the World Wide Web, Boston, December 11-14, 1995.
Concept mapping, semantic networks, hypertext, information modeling, and conceptual indexing all exemplify knowledge representation, in somewhat different ways.
Concept mapping seems to be rooted primarily in educational techniques for improving understanding, retention, and as an aid to writing.
A concept map is a picture of the ideas or topics in the information and the ways these ideas or topics are related to each other. It is a visual summary that shows the structure of the material the writer will describe. [p. 157]
- Thomas L. Crandell, Ph.D., Naomi A. Kleid, Ph.D., Candace Soderston, Ph.D. "Empirical Evaluation of Concept Mapping: A Job Performance Aid for Writers" Technical Communication, May 1996 (2nd quarter), pp. 157-163.
Semantic networks are often closely associated with detailed analysis of texts and networks of ideas. One of the important ways they are distinguished from hypertext systems is their support of semantic typing of links -- for example, the relationship between "murder" and "death" might be described as "is a cause of." The inverse relationship might be expressed as "is caused by."
Semantic networks are a technique for representing knowledge. As with other networks, they consist of nodes with links between them. The nodes in a semantic network represent concepts. A concept is an abstract class, or set, whose members are things that are grouped together because they share common features or properties. The "things" are called instances of the concept. For example, Femur is a concept representing the set of all femurs in the world; John Smith's left femur is an instance of the concept Femur.
Links in the network represent relations between concepts. Links are labeled to indicate which relation they represent. Links are paired to represent a relation and its inverse relation. For example, the concept Femur is related to the concept Upper Leg with the relation has-location. The inverse of has-location is the relation location-of, which relates Upper Leg to Femur.
- Edward Pattison-Gordon, "Knowledge Representation for Medical Concepts"
Representative examples of semantic network applications include ATLAS/ti and Thenetsys .
And if you're an InfoCentral personal information manager fan, you can use its object-oriented model to do many of the things you might do with a semantic network. InfoCentral is now sold by Corel Corp. .
Hypertext, known to most people these days by its implementation in the World Wide Web, is sometimes described as a semantic network with [substantial] content at the nodes. But the content itself -- the traditional document model -- seems to be the driving organizational force, not the network of links. In most hypertext documents, the links are not semantically typed, although they are typed at times according to the medium of the object displayed by traversing the link.
See also distributed hypertext systems.
Information modeling interests itself in precise specification of the meaning in a text, and in making relationships of meaning explicit -- often with the objective of rapid and accurate development of new software applications for business requirements.
Some of the essence of information modeling is expressed in the proceedings of a recent workshop on object-oriented systems:
How do we: elicit requirements from domain experts, formulate a complete and precise specification understandable to both domain experts and developers, and refine it using existing (or possible) implementation mechanisms.
Fuzzy terminology results in fuzzy thinking: precise and explicit definitions are essential for understanding and reuse . . . [p. vi]
- Proceedings of the Workshop on Semantic Integration in Complex Systems: Collective Behavior in Business Rules and Software Transactions, 16 October 1995. (Fourth Workshop on Specification of Behavioral Semantics). Institute for Information Management and Department of Computer and Information Systems, Robert Morris College, Coraopolis and Pittsburgh, Pennsylvania.
Conceptual (or "back-of-the-book") indexes are rarely discussed in the same breath as hypertext, conceptual maps, and semantic networks -- perhaps because indexers themselves sometimes relish the aura of "black art" surrounding indexing -- but the connection is fundamental. Conceptual indexes traditionally map key ideas and objects in a single work:
An index is a structured sequence -- resulting from a thorough and complete analysis of text -- of synthesized access points to all the information contained in the text. The structured arrangement of the index enables users to locate information efficiently. [p. 4]
- Nancy C. Mulvany, Indexing Books, University of Chicago Press, 1994
The organization of topics into parent-child, synonym, and "see also" relationships is a critical part of an effective professional index. Good indexes are not flat lists of names and ideas.
The terms knowledge sharing and information sharing are often used in conjunction with discussions of ontologies and knowledge representation. In the context of the following quote, the primary concern of information sharing is precision of expression and access in order to meet the objective of rapid product development.
Information sharing and decision coordination are central problems for large- scale product development. This paper proposes a framework for supporting a knowledge medium [stefik86]: a computational environment in which explicitly represented knowledge serves as a communication medium among people and their programs. The framework is designed to support information sharing and coordinated communication among members of a product development organization, particularly for the tasks of design knowledge capture, dynamic notification of design changes, and active management of design dependencies. The proposed technology consists of a shared knowledge representation (language and vocabulary), protocols for foreign data encapsulation and posting to the shared environment, and mechanisms for content-directed routing of posted information to interested parties via subscription and notification services. A range of possible applications can be explored in this framework, depending on the degree of commitment to a shared representation by participating tools. A number of research issues, fundamental to building such a knowledge medium, are introduced in the paper.
- T. R. Gruber, J. M. Tenenbaum, & J. C. Weber. "Toward a knowledge medium for collaborative product development." Proceedings of the Second International Conference on Artificial Intelligence in Design, Pittsburgh, pages 413-432. Kluwer Academic, 1992. See downloadable versions.
Contrast knowledge sharing with computer-supported collaborative work (CSCW).
Many workers in medical informatics are seeking to reuse knowledge in new applications and to share encoded knowledge across software environments. Knowledge reuse involves many dimensions, including the reapplication of lexicons, ontologies, inference syntax, tasks, and problem-solving methods. Principal obstacles to all current work in knowledge sharing involve the difficulties of achieving consensus regarding what knowledge representations mean, of enumerating the context features and background knowledge required to ascribe meaning to a particular knowledge representation, and of describing knowledge independent of specific interpreters or inference engines. Progress in the area of knowledge sharing will necessitate more practical experience with attempts to interchange knowledge as well as better tools for viewing and editing knowledge representations at appropriate levels of abstraction.
- M. A. Musen. (1992). Dimensions of knowledge sharing and reuse. Computers and Biomedical Research, 25, 435-467. See the Web address http://www-ksl.stanford.edu/knowledge-sharing/papers/README.html
Metadata is simply information added to a document (or a smaller unit of information) that makes it easier to access and re-use that content. It's also referred to as simply "data about data." You'll find metadata in many different forms, including key words in a software help system, the document profile information attached to documents in a document management system, and the classification information in a library card catalog.
There are, of course, distinctions in how these various disciplines and technologies implement metadata -- in substance as well as in formalisms. But the value of metadata for critical information is widely accepted as a basic element of knowledge management implementations. In fact, the term metadata has become so popular that it recently merited its own IEEE conference.
There is a strong interest in metadata in the "geographic information systems" (GIS) community -- the one concerned with maps, not the technology for graphic representation of the location of corporate intellectual assets. Claritech's Elise Yoder observes that the "motherlist" for current work on Metadata seems to be "Metadata Resources" .
Computer-based ontologies -- formal, structured representations of a domain of knowledge -- are commonly associated with artificial intelligeznce technology, where they were originally designed to serve as the raw material for computer reasoning and computer-based agents.
(See also taxonomy, knowledge representation, and knowledge re-use.)
An ontology is a set of definitions of content-specific knowledge representation primitives: classes, relations, functions, and object constants.
- T. R. Gruber. Ontolingua: A mechanism to support portable ontologies. Stanford University, Knowledge Systems Laboratory, Technical Report KSL-91-66, March 1992. Revision. Web version: http://www-ksl.stanford.edu/knowledge-sharing/papers/README.html#role-of-onto
But ontologies are no longer just for computers. They are frequently about knowledge sharing, whether among computers or among people.
Ontologies are also crucial for enabling knowledge-level interoperation of agents, since meaningful interaction among agents can occur only when they share a common interpretation of the vocabulary used in their communications. Finally, ontologies are useful in many ways for human understanding and interaction. [p. 1]
- Adam Farquhar, Richard Fikes, Wanda Pratt, James Rice, "Collaborative Ontology Construction for Information Integration," Web address: http://www-ksl-svc.stanford.edu:5915/
Perhaps the "mother of all ontologies" is the massive Cyc ontology being assembled by Lenat and others at MCC.
The aim is that one day Cyc ought to contain enough common sense knowledge to support natural language understanding capabilities that enable it to read through and assimilate any encyclopedia article - i.e., to then be able to answer the sorts of questions that you or I could, after having just read that article, questions which neither you nor I nor Cyc could be expected to answer beforehand.
- From "The Evolution of the Cyc Methodology," Web address: http://www.cyc.com/tech-reports/act-cyc-134-90/sectionstar3_2.html#SECTION0002000000000000000
Building a traditional AI ontology for a domain of expertise is generally viewed as a difficult and time-consuming task that requires a high level of expertise. As a result, sharing the components of ontologies -- and enabling the transfer of information about entities and relationships among different AI ontology-building systems -- is a significant concern among the AI community. The ARPA Knowledge Sharing Effort addresses this problem. See an overview, "The Knowledge Sharing Effort" by Robert Neches.
