Data portraits : aesthetics and algorithms

Abstract

While interacting online, one generates a multitude of personal data trails, both textual and behavioral. The data portrait is a way to collect, condense and represent these information trails, which are often time consuming and tedious to find and grasp when read linearly across web pages or domains, into an easy, legible, and compelling visualization. This thesis presents various data portraiture techniques that generate both individual and collective portraits of users participating in online social media. The data used in generating the portraits are unstructured text and publishing timestamps of Twitter micro-blog posts, as well as aggregate RSS feeds from FriendFeed. The strategies for depicting people's online personas explored in this thesis focus on the compression, mapping and visual representation components of the visualization pipeline. The resulting portraits attempt to maintain a tight connection with the data, and be legible to viewers, but at the same time, venture to explore more expressive visual forms, and engage with the evolving aesthetics of cinematography, typography and animation.