Introduction
Building profiles of registered users of a web site, as in case of a portal, is of importance if it goes beyond collecting the obvious information the user is willing to give at the time of registration. The starting point of our project is the assumption that a community of users is registered on a web site and that for each user a profile is built. Such a profile contains two parts of data: one obvious, given directly by the user and one less obvious, inferred by the user’s behaviour during his visits on the site.
Our prototype Gugubarra implements ideas developed at the database group (DBIS) at the JWG-university Frankfurt. The focus is how to build, manipulate non-obvious user profiles. Special emphases are paid on a feedback mechanism and several methods of measuring its results. Gugubarra works as a test application on real data by using the IT-specialists’ community viewzone.org.

Building the NOP
A web site (or a domain) is a collection of web pages that are linked together within the site. Each page has a specific content (in our model this relates to a specific topic). The relevant data to be used in the algorithm is extracted from various log files generated by the web server.
Based on this data, the user’s navigation through the site is brought together with the corresponding topics and with help of an algorithm the non-obvious user profile (NOP) is generated. Therefore the NOP shows the path of the user through the site and the footprints he left there.

Including a Feedback Mechanism
To measure the accuracy of a NOP we use a feedback mechanism, where the user is asked directly to enter his preferences, e.g., by presenting the values of his NOP and asking him to verify or correct them. From the feedback given by the user we build up another profile called the feedback profile (FP). Now Gugubarra compares the NOP with the FP and by use of different attempts a derived profile (DP) is build as a base for further calculations.

Using the Results
There are several ways to use the calculated profiles, e.g. the grouping of similar users in a cluster or the projection of user’s behaviour in the future. Another possibility is to adapt the presented content to the user’s interests. An important keyword in this context is the personalisation. Main goal of our work and building non-obvious user profiles is to improve the services offered on a web site.

"Gugubarra" is the Aboriginal name for the Kookaburra bird.

We are open to cooperation. If you are interested to get in contact with the Gugubarra team, please send us an e-mail at:
gugubarra@viewzone.org