Predicting User Preferences - From Semantic to Pragmatic Metrics of Web Navigation Behavior

Document Type

Conference Proceeding

Publication Date


Find this in a Library

Catalog Record


This paper aims at demonstrating a particular method to extract relevant information from navigation data. Participants had to perform web tasks while their navigation behavior was recorded. Then, they had to rate a list of presents based on the question: "how much would you like to receive this present for your birthday?" The semantic similarities (calculated with Latent Semantic Analysis) between users' navigation paths and presents' descriptions were used as implicit estimates of users' preferences. Results show that users' explicit preferences can be accurately estimated based on users' navigation behavior. Consequently, we propose this method to be used in building user models for adaptive web applications.

Catalog Record