Visualizing Event Sequences as Oscillating Streams

Abstract

In this paper, we introduce a new method to visually represent sequence structure in data. Like other methods for visualizing temporal or ordinal data, the representation directly maps absolute time or relative ordering of events from left to right horizontally. Unlike other methods, it also accumulates subsequences of events into streams that oscillate up and down vertically. By interactively adjusting the number of steps between vertical reversals, one can rapidly switch perspectives to show variation in event densities over time (one step), overall patterns of event accumulation (all steps), or short-range patterns of event accumulation (in between). In between, the representation reverses stream direction every N steps, accentuating variations in event accumulation while at the same time preserving visual continuity. We present a user study that compares the stream representation to Dotplots. The study validates the readability of the representation for effective visualization of sequence information in text data. It also shows how pairing stream and Dotplot views outperforms both of them individually for some analysis tasks.