Customizing Semantic Profiling for Digital Advertising

Personalization is the new magic buzzword of application development. To make the complexity of today's application functionalities and information spaces ``digestible'', customization has become the new go-to technique. But while those technologies aim to ease the consumption of media for their users, they suffer from the same problematic: in the age of Big Data, applications have to cope with a conundrum of heterogeneous information sources that have to be perceived, processed and interpreted. Researchers tend to aim for a maximum degree of integration to create the perfect, all-embracing personalization. The results are wide-range, but overly complex systems that suffer from issues in performance and scalability. Starting from our own, similar experience, we argue for a personalization of personalization - a clear-cut customization of profiling by limiting the information maintained in the system to those elements that are crucially relevant for the target task. We demonstrate the concern based on an application developed in our laboratory: a system that profiles web users based on their browsing history, with the goal of personalizing digital advertisements.