Multi-Level Visualization of Interrelated Data Entities Sandra Gama INESC-ID and Instituto Superior Técnico, Universidade de Lisboa Av. Rovisco Pais, 1 Lisboa, Portugal sandra.gama@tecnico.ulisboa.pt Daniel Gonçalves INESC-ID and Instituto Superior Técnico, Universidade de Lisboa Av. Rovisco Pais, 1 Lisboa, Portugal daniel.goncalves@inesc-id.pt ABSTRACT Nowadays, electronic devices are part of our daily routines, resulting in information generation at virtually any time and context. Due to different styles of interaction, data pro- duced by human activities is not only in considerable quan- tities, but it is also extremely rich, which makes it difficult to manage and analyze. Visualization has the potential to overcome this limitation: not only is it an excellent means to display large quantities of information, but it also alleviates cognitive load associated with data interpretation. We cre- ated an interactive multi-level layered visualization, in which time may be represented sequentially through layers. Data entities are displayed as circles with size proportional to a particular data feature we need to highlight, allowing imme- diate comparison between entities. By selecting an entity, we may see, through visual connectors, all the interrelated entities over the different time layers. User tests have shown that our visualization makes important information imme- diately perceivable, in a way that is easy to navigate and analyze. Categories and Subject Descriptors H.5.2 [Information interfaces and presentation]: User Interfaces General Terms Human Factors 1. INTRODUCTION Human activities, with increasing use of electronic devices, generate high quantities of information in a variety of con- texts. Such amounts of information are often quite complex and difficult to interrelate and analyze. Visualization, as an excellent means to display large quantities of data alleviates cognitive load associated with information interpretation [2], may be the solution to this shortcoming. We created a vi- sualization that displays data items in a timely manner and allows interactive exploration, providing mechanisms for in- terrelating and comparing items. It is minimalistic and easy to navigate, allowing an effective analysis. We adopted a multi-level layered view, in which each level corresponds to a time frame, where data items are depicted. This visual- ization makes general information immediately perceivable and allows interactive manipulation for further exploration of patterns such as relationships among entities and detailed information on a given entity. In section 2 we present our visualization and in section 3 we discuss user evaluation results, drawing several conclusions from our study. 2. APPROACH We created an interactive visualization, consisting of a main area and several additional elements, as depicted in Figure 1. The main area is composed of a multi-level interactive lay- ered visualization, in which each level corresponds to a time frame. Data entities are displayed as circles with size pro- portional to a particular feature that it verifies. We provide the means to represent the positive and negative aspects of a particular feature, or both: the entity circle is sub-divided into two semicircles displaying information both on positive and negative aspects through western conventional positive- negative color coding [2]. The leftmost, green, semicircle represents the positive aspect and the rightmost, red, semi- circle, displays the negative aspect of the feature. Semicircle size is proportional to the magnitude of both the positive and the negative aspects. This mechanism makes important information instantly perceivable. In fact, it makes it possi- ble to promptly understand the entities for which there are only positive (or negative) aspects to a given feature and also allows the immediate comparison between entities, without requiring further data exploration. Moving the mouse over a circle highlights it by assigning more saturation to its present color hue, making it more noticeable, and displays information on entity relationships. The current entity is linked to all the entities with which it has any type of interrelation through visual line connectors. Line connector thickness is proportional to the interrelation strength between two entities (which may be defined as a specific feature such as entity similarity): the thicker the line, the stronger the relationship is. Color is assigned to each interconnection individually, in order to avoid visual confusion and allowing immediate line discrimination. We did not use fully saturated colors, in order to keep our visual artifacts from competing for the user’s visual attention [2].