KN-AHS: An Adaptive Hypertext Client of the User Modeling System BGP-MS1

Alfred Kobsa, Dietmar Müler, Andreas Nill

Department of Information Science, University of Konstanz P.O. Box 5560-D73, 78434 Konstanz, Germany {kobsa,mueller,nill}@inf-wiss.uni-konstanz.de


Since hypertext is frequently read by users with differing knowledge and experience levels, it may at the same time be too difficult and too detailed for laypersons, and too redundant for experts. Boyle & Encarnacion [1994] showed empirically that an automatic adaptation of hypertext to the user's state of knowledge significantly improves text understanding as well as partially improving search speed. The system KN-AHS2 deals with this problem and adapts hypertext objects to the current user's state of knowledge. In contrast to other adaptive hypertext systems, the realization of KN-AHS took advantage of existing software products. TOOLBOOK [Asymetrix 1989], a widely available hypertext shell system, offered us a powerful tool for the implementation of the hypertext and its user interface. The user modeling shell system BGP-MS3 offered a wide variety of representation and inference possibilities that ensure flexible adaptation. Both tools run as independent software systems on a PC platform and interact via inter-process communication.

Observations based on user actions will be reported to BGP-MS by the hypertext system. The application can ask BGP-MS questions about the user and BGP-MS can in return report its current assumptions concerning the user.

KN-AHS draws assumptions about the userıs knowledge based on two information sources: namely an initial interview, and some of the hypertext actions which the user may perform. In the initial interview, questions are posed to the user that refer to his membership in clearly separable user subgroups (like 'computer science student'), and his exposure to PCs, hypertexts, etc. The user's replies become communicated to BGP-MS, which can activate initial stereotypes for the user. Certain actions that the user may perform at the interface of the hypertext system give rise to assumptions about his familiarity with individual concepts:

If the user requests an explanation, a graphic, an example or a glossary definition for a hotword, then he is assumed to be unfamiliar with this hotword. If the user unselects an explanation, a graphic, or an example for a hotword, then he is assumed to be familiar with this hotword. If the user requests additional details for a hotword, then he is assumed to be familiar with this hotword. With each hotword for which more information can be requested, a SB-ONE4 concept that represents this technical term in BGP-MS is associated. When KN-AHS draws an assumption about the user's familiarity with a hotword, KN-AHS notifies BGP-MS that the corresponding concept is known or unknown to the user. KN-AHS deals with stereotypes by using the stereotype managing mechanism of BGP-MS. This mechanism analyzes observations received from KN-AHS and checks the activation and retraction conditions of all stereotypes. It will then enter inheritance links between the individual user model and those stereotypes that become active, and delete links to stereotypes that become deactivated. In the current domain of KN-AHS, more than one stereotype can be active at the same time. Inferences within KN-AHS are based on the observed assumptions and technical domain knowledge represented in a SB-ONE concept hierarchy. When the user switches to a new text object, KN-AHS aims at adapting it to the user's presumed conceptual knowledge. For each hotword in the new text object, it asks BGP-MS about the user's familiarity with the corresponding SB-ONE concept. The hotword is then treated in the following way:

If the user is unfamiliar with the associated concept, an explanation gets automatically added to the hotword. Also, an icon that symbolizes an available graphic for the hotword is placed near the hotword. If the user is familiar with the hotword, more details are automatically added after the hotword. If no information is available from BGP-MS concerning the userıs familiarity with the hotword, then the hotword is not changed.

References

Asymetrix Corporation 1989. Using TOOLBOOK: A Guide to Building and Working with Books (Version 1.5). Washington.

Boyle, C., and Encarnacion, A. O. 1994. An Adaptive Hypertext Reading System. User Modeling and User-Adapted Interaction. In press.

Kobsa, A. 1990. Modeling the User's Conceptual Knowledge in BGP-MS, a User Modeling Shell System. Computational Intelligence 6:193-208.

Kobsa, A. 1991. Utilizing Knowledge: The Components of the SB-ONE Knowledge Representation Workbench. In: J. Sowa, ed.: Principles of Semantic Networks: Exploration in the Representation of Knowledge. San Mateo, CA: Morgan Kaufmann.

Kobsa, A., and Pohl, W. 1994: The User Modeling Shell BGP-MS. User Modeling and User-Adapted Interaction. Submitted.

Profitlich, H. J. 1989. SB-ONE: Ein WissensreprŠsentationssystem basierend auf KL-ONE. Master Thesis, Dept. of Computer Science, Univ. of SaarbrŸcken, Germany.

  1. This work was supported by the German Science Foundation (Grant Ko-1044) and the Univ. of Konstanz (Grant AFF 17/92).
  2. KN-AHS stands for KoNstanz Adaptive Hypertext System.
  3. BGP-MS stands for Belief, Goal and Plan Maintenance System [Kobsa 1990, Kobsa & Pohl 1994].
  4. SB-ONE is of the knowledge representation languages used within BGP-MS. It belongs to the family of KL-ONE type languages [Profitlich 1989, Kobsa 1991].