Myrror: a platform for holistic user modeling: Merging data from social networks, smartphones and wearable devices

In this article, we present a platform that allows the creation of a comprehensive representation of the user that we call a holistic user model (HUM). Such a representation is based on the intuition that users’ personal data take different forms and come from several heterogeneous sources. Accordingly, we designed a pipeline that: (1) extracts personal data from three examples of important classes of such sources, namely social networks, wearable devices and smartphones; (2) processes these data through natural language processing and machine learning techniques; (3) stores the output of such processing in a user model that encodes different aspects of people’s life, such as demographic data, interests, affect values, social relations, activities and physical states. The resulting representation is made available to the user and to external developers. In the first case, a web interface allows the user to browse through her own personal data and to consult different facets of her HUM, in order to improve her self-awareness. In the latter, holistic user profiles are exposed through a REST interface and can be exploited by third-party applications to provide personalized services based on HUMs. In the experimental session, we evaluated usability and acceptability of the HUM in a user study which investigated how people were willing to use it. The results confirmed the effectiveness of our design choices and built the foundations for future usage of these profiles in personalized applications.