“Multicellular” Visualization      “Multicellular” Visualization      “Multicellular” Visualization     
DESIGNER
Muxing Chen
KEYWORD
#Visualizing Complexity
#Visual Composition
#Composition Framework
#Composite Visualization
ADVISOR
Paolo Ciuccarelli

Complex datasets (such as heterogenous, large-scale, spatiotemporal data, etc.) are becoming more increasingly accessible, as we are able to generate, collect and store greater quantities and varieties of information. However, no single visualization is optimal for all possible tasks. To address this challenge, a strategy called composite visualization combine multiple single visual representations in the same visual space in order to display different aspects of the complex dataset.


image
Fig 1. 100 Years of London Performance by Guillermina Sutter Schneider

Composite visualization consists of multiple component visualizations and a set of relations between them. This thesis project focuses on developing a composition framework that identifies ten composition patterns, based on the relations between component visualizations and with which we can classify and deconstruct composite visualization.


The thesis starts by contemplating what are the fundamental relations between component visualizations.

image
Fig 2. Concept map of composite visualization and the framework developed in this thesis

Fundamental Relation
Spatial Relation
image

Visual composition of two graphic representations or combining one with a group of visualizations in the same geometrical space inevitably requires consideration of the spatial relation between them.


The figure above depicts five types of spatial relations between shapes in a two-dimensional geometrical plane, which are included, inscribed, intersected, circumscribed, and separated. In the context of composite visualization, the intersection between two components is relatively rare, so this thesis does not account for this circumstance. Layering pattern is defined when the spatial relation between two visualizations is included or inscribed, whereas integration pattern is defined when the spatial relation between two visualizations is circumscribed and separated.


Fundamental Relation
Data Relation
image

Another fundamental relation is data relation. Due to the purpose of combining multiple visual representations to show the different facets of the dataset, there must be some inherent data relations between the visualizations rather than mechanically bringing them together. This thesis specifies five types of data relation between visualizations, which are listed above.

Composition Framework
image
Fig 3. Ten composition patterns defined by spatial relations and data relations between two visualizations

With two spatial patterns and five types of data relation specified, this thesis formulates a framework that is a two-by-five matrix that identifies ten composition patterns, as demonstrated in the figure.

Composition Framework
Item-to-item Layering
image
Fig 4. Brain Drain by Accurat

Item-to-item layering pattern is defined as one visualization overlays on another visualization that shares related data items. The above composite visualization is a typical example of this pattern, and one way to deconstruct it is to break it down into a scatter plot, a network diagram, and data glyphs. (Click to view more information about composition patterns and corresponding examples)

Physical Artifact
image
Fig 5. Front view of the installation
image
Fig 6. Side view of the installation

Inspired by Exploring the Brain Drain, this exhibited physical artifact reveals the ambiguity of composite visualization deconstruction. Three sets of acrylic sheets display three possible ways to deconstruct the same composite visualization (as seen from the front). Visual elements distribute differently on three layers of each set (viewed from the sides). There is currently no consensus and guidelines on how to deconstruct composite visualizations, which makes communicating the deconstruction of composite visualization equivocal. For example, this visualization can also be broken down into a bubble chart, a network diagram, and data glyphs.