Magazine article AI Magazine

The Fifth International Conference on User Modeling

Magazine article AI Magazine

The Fifth International Conference on User Modeling

Article excerpt

Almost 100 attendees with backgrounds in AI, human-computer interaction, psychology, education, library sciences, and other related areas gathered in Kailua-Kona, Hawaii, in early January 1996 for paper sessions, invited talks, tutorials, miniworkshops, special interest group (SIG) meetings, system demonstrations, a poster session, and a doctoral consortium. The invited talks included "Multimodal Interactive Maps: Designing for Human Performance" by Sharon Oviatt (Oregon Graduate Institute) and "Moving Up the Information Food Chain: The Internet Softbot and the METACRAWLER" by Oren Etzioni (University of Washington). Tutorials included "Student Modeling and Intelligent Tutoring Systems" by James Greet (University of Saskatchewan, Canada), "Bayesian Networks, Dempster-Shafer Theory, and Fuzzy Logic in User and Student Modeling" by Anthony Jameson (University of Saarbrucken, Germany), and "Multimedia and User Modeling" by Mark Maybury (MITRE). Miniworkshops included "The Commercial Potential for User Modeling," "Standardization of User Modeling Shell Systems," and "User Modeling for Information Filtering on the World Wide Web." SIG meetings included "Developing Adaptive Applications (with the User Modeling Shell System BGP-MS)."

Two papers shared the best paper award. Kathleen McCoy and Christopher Pennington (both from the University of Delaware) presented a system for correcting English syntax using malrules. A subset of likely malrules is selected for particular users based on a model of the user's first language and a model of second-language acquisition that takes into account language transfer problems. They applied their method to teaching written English to American Sign Language speakers. Sharing the best paper award was Susanne van Mulken (University of Saarbucken), who presented a method for probabilistically assessing the ease of understanding of interactive, computer-generated, multimedia presentations. She used Bayesian networks to combine the knowledgeability of the user (in the domain and in graphic conventions) with presentation factors such as the degree of ambiguity of a multimedia convention (for example, zigzag lines) in context and the familiarity of the convention as represented by its frequency of use. The Bayesian network (based on work by Anthony Jameson) allows updating of both user knowledgeability and presentation factors on user feedback.

Paper sessions were entitled Information Filtering, Adaptive Interfaces: Principles and Techniques, Student Modeling, Plan Recognition, Applications, Natural Language, and Cognitive Models. The Information-Filtering session included papers about agents to support search on the World Wide Web (for example, looking for new links, notifying about updates to links, and having uniform access to search engines); the use of analogical user modeling (analogy over a conceptual granularity hierarchy) for document filtering; a comparison of movie-selection techniques using features (predict movie rating from similarity of the movie's features to previously rated movie features) or cliques (predict rating based on the average rating of the user's clique or group); and information filtering using a combination of vector-space model, document classification by unsupervised learning, and a relevance-feedback user model.

The paper session on adaptive interfaces featured presentations on learning the user's idiosyncratic grammar using an adaptive parser within a unification-based grammar formalism, an investigation of the effect of the style of graphic data models on the user's reading strategies, and the use of user models to design presentations in real time with immediate user feedback in a formal abductive framework. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.