Semantics happen: knowledge building in spatial hypertext

Abstract

Hypertext represents ideas through chunks of text or other media interconnected by relations, typically navigational links. The similarity to knowledge representations such as frames and semantic nets has led to much effort in using hypertext systems for knowledge representation and extending hypertext systems to make them able to express more. This work has met with limited success due to difficulties including the tacit and situated nature of much knowledge. Instead of viewing knowledge expression as an all at once event, we view it as a constructive process, i.e. knowledge building. The Visual Knowledge Builder (VKB) lets users express content via visual or textual means and later formalize that content in the form of attributes, values, types, and relations. VKB proactively supports this process through a set of suggestion agents whose interaction with the user is mediated by the suggestion manager. Preliminary evaluation of the suggestion manager and suggestion agents yields positive results but further confirms that there is no "silver bullet" for knowledge engineering -- semantic expression is most likely to happen during, and is driven by, task performance.