WEB TECHNOLOGIES

Using Knowledge Anchors to Reduce Cognitive Overhead

Stephen Ransom, Manx 5 Pty Ltd. Xindong Wu, University of Vermont

incorporating some degree of layering to lower visual complexityor the routes leading from the current frame, but these efforts have thus far proved ineffective.

KNOWLEDGE ANCHORS
We believe that a more practical and effective approach is to use bidirectional knowledge anchors. These are points within a frame that have significance to the authorand later to the reader. A knowledge anchor is the point within the frame from which the user can trigger a link. An anchor designates either an area from which a reader may want to branch to investi-

eb publishing offers a major advantage over traditional print-based methods because it can link to external documents or other places within a lengthy document. Hypertext can potentially connect to any information on the Internet, including both data and executable programs. Thus, links not only let authors create dense, complex works that would be difcult to read in a linear fashion, they also let them incorporate into their online documents a rich array of multimedia such as audio and video. Existing Web development techniques rely on static links that map back to an existing pointer or index entry in the underlying system. Such hard links imply the existence of only one correct relationship for any given hypertext document and all potential users. The author alone is thus logically obligated to insert and modify every anchorthe point within a frame from which the user can trigger an internal or external link to explore a new subject or expand on the current one. The author must also determine how these changes affect all other existing anchors to determine whether new links should be inserted in either direction, a task that grows exponentially as the author proceeds.

Knowledge anchors facilitate the process of developing and updating hypertext documents.
COGNITIVE OVERHEAD
These demanding requirements result in two forms of cognitive overhead, a well-known Web design problem that Jeff Conklin dened as the additional effort and concentration necessary to maintain several tasks or trails at one time (Hypertext: An Introduction and Survey, Computer, Sept. 1987, pp. 17-41). This leads to anarchy at the macro level: Following the same path through the Web usually leads to different outcomes, and too often links return a Forbiddenyou dont have permission to access or Document not found error message. A document that is available one day may be gone or completely rewritten the next, thus destroying its relevance to the user. To control versioning, authors restrict jumps to their own material, which severely limits their Web sites potential value. In response, researchers have developed maps that show either an overview of the entire systemusually gate other related material or an area to which a reader may want to jump. A typical anchor, implemented as an anchor element <A> in HTML, only links outward to other anchors. A knowledge anchor contains two types of concept lists. The inward list describes the anchors immediate surroundings and why a reader might want to jump there. An anchor at a point of general interest could have many such concepts, while one intended only as a transition point might have none. The outward list describes what a reader might seek jumping from that point to another. This list could be empty if the author does not see the anchor as a logical branching point in the document. The author provides both sets of characteristics based on knowledge of the documents content and target audience, and can use any terms for anchors interconnecting his own material. Each knowledge anchor polls other anchors inward lists for links that match its outward list entries. An
November 2002

111

W e b Te c h n o l o g i e s

omputer Wants You

Computer is always looking for interesting editorial content. In addition to the articles we publish in our theme issues and special issues, we have other feature sections such as Perspectives, Computing Practices, and Research Features as well as numerous columns to which you can contribute. Check out our author guidelines at http://www. computer.org/ computer/author. htm for more information about how to contribute to your magazine.

Steps to Construct Knowledge Anchors

Once an author has created a frame, constructing a knowledge anchor only requires following a few simple steps: Identify all of the knowledge anchors within the framenew areas where the reader may want to explore or expand on the current one. Create the inward list of concepts describing the anchors context (or why a reader might like to jump there). Create the outward list of concepts describing material that the author believes the reader might want to see. Use an automated tool such as our Maigh hypertext system (S. Ransom, A Hypertext System Addressing the Issues of Disorientation and Cognitive Overhead, doctoral dissertation, Dept. Software Development, Monash Univ., Victoria, Australia, 1996.) to match anchor requests and anchor responses. Verify that the anchors have generated appropriate links.

author can use knowledge anchors to dynamically generate and store internal, one-way soft links within the requesting anchor for as long as the initiating frame requires. An author creates soft links dynamically as required, and they exist for as long as the initiating frame requires them. From the resulting possible matches, the author can choose and annotate those links that t the desired hypertext framework, as outlined in the Steps to Construct Knowledge Anchors sidebar. Alternatively, the computer can identify potential links according to a preselected degree of congruency, thus lowering cognitive complexity. The resulting list, otherwise equivalent to a traditional hypertext node map on the frame, uses only local links. To incorporate the most up-to-date links, authors traditionally insert anchors after nearly completing the base hypertext, which requires familiarity with the entire work. In contrast, knowledge anchors encapsulate only local hypertext information, thus authors can insert them on the y during development, greatly facilitating the process of selecting and identifying related concepts. This approach also makes simultaneous development of the entire hypertext unnecessary.

ypertext authors can use knowledge anchors to label all points of interest in their work without having to know exactly what relevant information exists on the Web at a given time. In the case of large Web sites, design teams can collectively come up with terms conceptually related to the sites content. These terms would constitute the knowledge of our proposed bidirectional anchors as well as a metadescription of the document to strengthen search engine results. However, effective matching can only occur when readers use the same terms. I Stephen Ransom is director of Manx 5 Pty Ltd., a financial services company based in Victoria, Australia. He is a member of the ACM. Contact him at ransom@melbpc.org.au. Xindong Wu is professor and chair of the Department of Computer Science at the University of Vermont. He is a member of the IEEE Computer Society, the AAAI, and the ACM. Contact him at xwu@emba.uvm.edu.
Editor: Sumi Helal, Computer and Information Science and Engineering Dept., University of Florida, P.O. Box 116125, Gainesville, FL, 32611-6120, helal@cise.u.edu

112

Computer