Estimating a user's degree of interest in a page during web browsing

Abstract

This paper describes the relationship between, on the one hand, operations performed by a user while browsing a Web page that seem to indicate interest in the content and, on the other hand, the user's real interest or lack of interest in that page. One way of acquiring information from a mountain of information is to use a personalization service that provides information tailored for each user. The system on which such a service is based has to learn what a user is interested in. We examined some user operations that seem to indicate interest. The results of our investigation can be used in the system in personalization service to learn about a user's degree of interest. In a preliminary survey of users' operations during Web browsing, we observed five users while they were browsing the Web and asked 26 users to complete a questionnaire about their operations during Web browsing. This preliminary survey clarified which operations indicate interest. We discuss a prototype system that detects DOM (Document Object Model) operation events in a Web browser and extracts interest-indicating operations from operation events. The prototype also estimates the user's degree of interest in a page. To evaluate the prototype, we carried out an experiment, in which 15 users participated. The results of this experiment show (1) the extracted interest-indicating operations are related to the user's actual degree of interest and (2) the estimated degree of interest to a page also reflects the user's actual degree of interest to some extent.