Personalized systems are systems that maintain a model of the user and use this model for adapting themselves to the individual user. Recommender systems are the most well-known type of such systems. User Modeling is the process of building this model or – in other words – of getting to know the user. User Modeling is a cross-disciplinary research topic that can be studied from different perspectives and disciplines: from human-computer interaction to artificial intelligence, from psychology to philosophy, from design to linguistics.
After a brief overview of these and other ways of conceiving the field of User Modeling, the tutorial will mainly focus on data-oriented and reasoning aspects of User Modeling. User Modeling methods try to create digital representations of users, and then use these models for calibrating and adapting the interface or the content based on these models.
Methods, techniques and approaches of User Modeling changed along with the changes in the Web, as new opportunities and challenges have arisen. In the recent years, with the advent of the Web 2.0 and the growing impact of the Internet on our every day life, people start to use more and more different Web applications. They manage their bookmarks in social bookmarking systems, communicate with friends on Facebook and use services like Twitter to express personal opinions and interests. Thereby, they generate and distribute personal and social information like interests, preferences and goals. All these information can enrich the user model with more data about users and their interests and preferences. As a result, early methods that were hand-crafted and knowledge-based have been largely replaced by statistical approaches, as known from the fields of machine learning and data mining.
At the same time, the growing popularity of Semantic Web techniques allowed for ontology-based user models, which are able to build smarter content-based systems, showing to be very effective in solving some drawbacks of recommendation methods, such as the cold start problem and data sparsity. Moreover, the Linking Open Data (LOD) project helped to produce billions of RDF statements that are now published on the Web. This Web of Data, with its enormous availability of such free machine-readable knowledge, opens new possibilities for User Modeling: it allows to enrich the knowledge of the user by extending it with more details on items that users are interested in or have listed in their profile and by reasoning using these additional details. As a basic example, a simple statement that a user lives in Berlin could be extended with all data from DBpedia on this city and compared with related cities. This allows for types of reasoning that increase the number of possible items to recommend to user, increase the precision and the recall of the recommendation, increase the diversity in recommendation results and allow for cross-domain recommendation.
Starting from these premises, the tutorial will focus on describing the User Modeling process and the role of the Social and Semantic Web in the enrichment of user model, in relation to coverage and precision of the user model and effectiveness of recommendation.