Profiling user interests on the social semantic web

Abstract

The World Wide Web is evolving towards an ecosystem of applications and services offering personalised content to its users. At the same time, the widespread adoption of social media led its users to provide portions of their personal data on several different services for socialisation or personalisation purposes. The automated extraction of users' interests from personal Social Web data is becoming an essential part of the current Web applications for personalisation and recommendation. Such personalisation is required in order to provide an adaptive Web to users, where content fits their preferences, background and current interests, making the Web more social and relevant. Current techniques of personalisation systems analyse user activities on a social media system and collect sets of tags, entities or links to represent users' interests. These sets representing users' interests, also called user profiles of interests, are often missing a deeper "understanding" of the represented interests. Moreover, these user profiles cannot be easily exchanged between social media systems, therefore lacking portability and interoperability of personal user information. As a remedy, we propose a complete methodology for profiling user interests that leverages Semantic Web technologies and provenance of Social Web data. The Semantic Web represents a prominent recent approach attempting to provide the Web with a meaning not only people, but also machines can process. We adopt Semantic Web technologies for creating a standard interoperable representation of user profiles of interests. This allows for aggregation of heterogeneous user models from different social websites, and knowledge enrichment about user entities of interest. Moreover, we leverage provenance management of Social Web data to retrieve complete information about data producers (either applications, software agents or users) and increase the accuracy of user profiles. Provenance of data can be considered as one of the core building blocks for establishing data quality measures, for enhancing the knowledge acquisition/filtering process, and the user profiling phase. We investigate and evaluate a set of heuristics for mining users' interests from their social activities on heterogeneous social media websites and propose different approaches and measures for aggregating, enriching and ranking users' concepts of interest. Finally, we evaluate our methodology for profiling user interests in a practical Web personalisation scenario.