Abstract
The increasing plethora of information available on the Web necessitates effective personalization mechanisms that allow users to retrieve pieces of information that are more likely to be of interest to them. Fortunately, the democratization of the Web (aka. Web 2.0) that provides online users with widely available tools that allow them contribute and integrate themselves into the global information space not only increases the sheer amounts of data, but also offers opportunities to extract semantic meaning. This paper therefore presents an integrated approach that harnesses geo-tagged web resources like tourism services or track data from bike trails to derive semantic annotations for objects from their geographic proximity. Following this, a recommendation mechanism is proposed that hybridizes collaborative mechanisms with the additional knowledge about semantic annotations to make predictions about what will be relevant to a user in a specific situation. The utility of this integrated approach is showcased by an adaptive Web-GIS scenario that supports travelers in their decision making. Finally, the proposed algorithms are evaluated using historical log data from real users who were exploring the map of a tourism destination. The results indicate that, despite very short interaction sequences, improvements compared to a collaborative filtering baseline can be achieved. An additional advantage lies in offering users more detailed options to express their search preferences that is not quantified by the presented evaluation scenario.