2. Broad motivation
Proliferation of digital data (esp. geospatial and
network)
The rise of ‘big data’, coupled with increasing interest in
social media trace data
Many static, image-based visualizations being
produced, with many tools geared up for producing
them…
However, static images loose depth & sacrifice
complexity and nuance
We concluded there is a lack of elegant, easy-to-use
tools for non-technical researchers to create
interactive visualizations
3. Existing tools, libraries
That is not to say that some tools do not
already exist! For example:
Many Eyes
Google Charts, Maps
Simile Widgets
Sigma.js, D3.js
Miso
Raphael, Dojo Toolkit
oCanvas
Processing
4. Why interactive?
What are the limitations of conventional (static) approaches?
Logic of conventional visualization is a product of the limits of traditional
paper-based dissemination methods. Cannot be interactive, by
definition.
This reduces the intelligibility of complex data sets (by
requiring inelegant and confusing visual representations)
…or forces simplification, resulting in a loss of explanatory
power.
Multidimensionality of data (esp. ‘Big Data’) needs dynamic
representation
This encourages user/reader exploration
Allows finding specific values without returning to source data
Promotes richer dissemination of findings outside of (and in addition to)
traditional academic publication
5. Project development parameters
Browser native and standards compliant code
No flash, no java, no proprietary formats
Standalone/Offline
Can work in ebooks
Can work on tablets even if disconnected
Easy to use
Online interface to take the researcher “from
datafile to visualization”
6. What types of visualizations?
Network diagrams
World choropleth maps
(Spatial treemaps)
7. Networks
Network data increasingly common
Structural explanation gaining momentum within
social science
Influx of tools and techniques from other
disciplines (Comp Sci/Physics)
SNA community has an historical aversion to
disseminating raw data
A perceived complexity is seen to prevent the ‘lay’
interpretation of network data.
Nevertheless, visualization a crucial
analytical/interpretive step.
8. Implementation Considerations
Computational Demands
Hundreds/thousands of nodes to be rendered. Force directed
layout algorithms scale poorly in javascript, and are difficult to
implement. Interactivity adds further demands.
File Formats
Many to choose from (graphml, gexf, gdf, dl, net). Some XML,
some not. No clear winner.
Rendering Technology
SVG, Canvas, WebGL…and the complex relationship between
them.
Many Network Types – One (Extendable) Interface
Need to avoid overly specific features/becoming a ‘twitter
visualizer’.
9. Implementation Details
• Sigma.js as the renderer
• Renders to canvas element using its own internal drawing
methods and scene management (layers for nodes, edges,
etc.)
• HTML5/CSS3/JS interface, jQuery for transitions and
effects
• Clean and customisable framework, based on a
single json configuration file
• Editable by hand (for advanced users)
• Parser to turn common file formats into pure
javascript data structures (json)
• Increases rendering speed
• Preserves existing layout (so no layout needed in browser)
16. Issues/Limitations
No web based wizard (time constraints)
Upper bound to scalability due to rendering
method
Potential to implement WebGL
Dependence on existing skills
Graph layout
Framing
Software packages (Gephi)
24. Implementation
SVG/Raphael
Lightweight, standalone, tablet-friendlish
Abstracted json configuration
config in one place, human readable
setup wizard only needs to write two files (data,
config)
26. Issues/Limitations
Wizard UI wanting
Hard to mash up with GIS shapefiles
Unable to having backing/tile map
Possibly extend with leaflet.js implementation,
but issues with speed portability.
27. Thanks!
Use the tools / contribute (all code is open
source)
Sign up for our mailing list; subscribe to RSS
feed
Extend the spirit of the project! Create your
own class of visualisation, using open
standards, and common data formats.
http://blogs.oii.ox.ac.uk/vis