The tutorial will be organized in three parts.
FIRST PART. The tutorial will start with providing a brief historical overview on the research in the area of User Modeling, focusing on how methods, techniques and approaches have changed following the changes in the Web. From the traditional Web to the Social Web, the Semantic Web and the Web of Data, we will describe the historical evolution of User Modeling, moving from explicit rule-based, stereotype or logic-based models to implicit, statistical models and probabilistic techniques that have become more common with the availability of user and usage data in the Web 2.0. At the end of this part, some state-of-the-art examples of adaptations and recommendations will be presented.
SECOND PART. After this overview, we will go deeper inside the data-oriented aspects of the User Modeling process, presenting methods and techniques for:
- obtaining user data, varying from explicit methods (such as profile creation, item rating, communication and commenting) to implicit methods (Web Usage Mining, crawling of social media, logging methods)
- enriching user data with user-related information. This includes obtaining information on users that a user is connected with (FOAF, Facebook, Twitter, …) and items that users are interested in or that they indicated in their profile (descriptions and data about movies, books, cities, public persons, to mention a few)
- representing user models (e.g., stereotypes, flat models, hierarchical models, ontological models, overlay)
- reasoning about them (inductive and deductive reasoning) and content-based and collaborative methods for providing recommendations.
- evaluating user models and recommendation with dataset-based methods, to show the impact of data enrichment on contextualized recommendations.
THIRD PART. The last part of the tutorial will be devoted to a practical exercise in order to experiment some form of profile enrichment, showing the impact on collaborative filtering, e.g. different neighborhoods.