Experiences with semantic net based hypermedia

The Many Using and Creating Hypermedia (MUCH) system is based on the Dexter model and treats the storage layer as a semantic net. The MUCH system provides a number of recommended link types for representing application domain concepts, such as thesauri, documents, and annotations. Users of the system are expected to use those link types in the course of authoring meaningful hypermedia. This paper is based on the logs of usage of the MUCH system over two years by over 200 people. Contrary to the expectations of the builders of the MUCH system, the users did not exploit the ability to type semantic links. Typically authors used the default link type regardless of their semantic intentions. When a link type other than the default type was chosen, that choice was often inconsistent with the way another user would label a similar link. The system has proven to be useful for authoring conventional documents. Authors, however, were not practically able to produce hypertext documents. Based on these experiences a new system, RICH (Reusable Intelligent Collaborative Hypermedia), has been designed and built which emphasizes rules for typing links and maintaining the integrity of the semantic net.