Understanding a user’s emotional state is becoming increasingly important in
the development of adaptive and intelligent mobile applications. This thesis
addresses the problem of inferring valence and arousal in mobile users based
on the text displayed on their smartphone screens. We developed a mobile application that uses the Experience Sampling Method (ESM) to collect
users’ self-reported emotional states during phone use, while simultaneously
capturing on-screen text. The collected data was used to train machine
learning models aimed at inferring the user’s emotional state. A user study
confirmed the feasibility of the approach and demonstrated the potential to
understand user affect based solely on the displayed content. The results lay
the groundwork for the development of emotion-aware mobile applications
and tools for psychological research in naturalistic settings.
|