Title :
Browsing and keyword-based profiles: a cautionary tale
Author :
Shepherd, Michael ; Watters, Carolyn ; Duffy, Jack ; Kaushik, Raj
Author_Institution :
Fac. of Comput. Sci., Dalhousie Univ., Halifax, NS, Canada
Abstract :
In the research presented, adaptive user profiles are used to rate Web pages with respect to possible user interest. The user profiles consist of weighted keywords and the adaptation is based on the Hebbian Learning Model with direct user feedback. A user study was conducted to determine if the system would "learn" over a series of five sessions when there was no explicit task other than to browse. The results are analogous to reading the news, i.e., it is not possible to predict what pages a user will read if there is no explicit task. We suggest that a shift away from content to document style (genre) and other user characteristics may be more effective.
Keywords :
Hebbian learning; feedback; human factors; information resources; information retrieval; user interfaces; Hebbian Learning Model; Web pages; adaptive user profiles; direct user feedback; document style; explicit task; keyword-based profiles; user characteristics; user interest; user study; weighted keywords; Adaptive filters; Computer science; Feedback; Hebbian theory; Humans; Information analysis; Information filtering; Lifting equipment; Strategic planning; Web pages;
Conference_Titel :
System Sciences, 2001. Proceedings of the 34th Annual Hawaii International Conference on
Conference_Location :
Maui, HI, USA
Print_ISBN :
0-7695-0981-9
DOI :
10.1109/HICSS.2001.926475
