dataspace data visualization david karger february 8, 2010 nsf site visit to mit dataspace 1
DESCRIPTION
DataSpace Visualizing Heterogeneous Data We propose to make data more available Which will encourage scientists to gather it across disciplines How will they look at it once they have it? – Discipline-specific visualization tools won’t know what to do with other Disciplines’ data We need data-agnostic visualization tools – Must we recourse to spreadsheets? February 8, 2010 NSF Site Visit to MIT DataSpace 3TRANSCRIPT
DataSpace
Data Visualization
David Karger
February 8, 2010
NSF Site Visit to MIT DataSpace1
DataSpace
Visualization Drives Insight
• We need visualizations to help us understand our data– Formulate hypotheses– Then test/confirm them
• We use visualizations to communicate our insights to others
February 8, 2010
NSF Site Visit to MIT DataSpace2
DataSpace
Visualizing Heterogeneous Data
• We propose to make data more available• Which will encourage scientists to gather it
across disciplines• How will they look at it once they have it?– Discipline-specific visualization tools won’t know
what to do with other Disciplines’ data• We need data-agnostic visualization tools– Must we recourse to spreadsheets?
February 8, 2010
NSF Site Visit to MIT DataSpace3
DataSpace
Visualization Ontologies
• We can ontologize the different ways of looking at information– Selection of attributes, layout, means of filtering
• View descriptions can travel with the data• Ontology-aware tools can generate suitable
visualizations on a variety of platforms• Scientists can publish their visualizations with
their conclusions
February 8, 2010
NSF Site Visit to MIT DataSpace4
DataSpace
Fresnel Visualization OntologyLongwell Browser
• dsp:publicationFacets rdf:type facets:FacetSet; facets:facets ( dc:type dc:publisher dc:contributor dc:subject ); facets:types ( ow:Publication ) . dsp:publicationLabelLens rdf:type :Lens; :classLensDomain ow:Publication; :group gr:group; :purpose :labelLens; :showProperties ( dc:title ) .dsp:publicationLens rdf:type :Lens; :classLensDomain ow:Publication; :group gr:group; :purpose :defaultLens; :showProperties ( dc:description dc:identifier dc:creator dc:contributor dc:date dc:subject dc:type dc:publisher dc:rights ) . dsp:creatorFormat rdf:type :Format; :group gr:group; :propertyFormat [ rdf:type :FormatDescription ; :contentNoValue "[none]" ];
February 8, 2010
NSF Site Visit to MIT DataSpace5
DataSpace
DataSpace
Exhibit Framework
• In-HTML-document visualization ontology• Author creates a data set– Excel spreadsheet, tabbed file, XML file, JSON
• Then writes HTML document– Specify view by embedding HTML “tags” from viz
ontology• Our javascript makes the data-interactive viz
February 8, 2010
NSF Site Visit to MIT DataSpace7
DataSpace
Deployment
• Deployed 2 years ago• Hundreds of exhibits created in the wild• Healthy open source community– Fixing bugs, extending with new views
• Timeline on opendata.gov for several months
February 8, 2010
NSF Site Visit to MIT DataSpace8
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace9
DataSpace
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace11
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace12
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace13
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace14
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace15
DataSpace
DataSpace
DataSpace
DataSpace
DataSpace
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace21
DataSpace
INSIDE A WIKI
February 8, 2010
NSF Site Visit to MIT DataSpace22
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace23
DataSpace
OR A BLOG
February 8, 2010
NSF Site Visit to MIT DataSpace24
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace25
DataSpace
Going Forward
• Refine our understanding of the universal concepts of information visualization
• Deploy domain-independent tools that can use visualization descriptions to create effective visualizations of arbitrary data drawn from many locations– Data comes with its visualization descriptions– Fed to specialized widgets that fit the descriptions
February 8, 2010
NSF Site Visit to MIT DataSpace26
DataSpace
What Widgets?
• Descriptions aren’t programs• Won’t implement highly specialized views– Gene map, pie chart, heat map, star chart, scatter
plot, chemical bond diagram…• But these views can be
implemented/packaged/downloaded for incorporation into the visualizers– Visualization “app store”
• Then referenced in viz description
February 8, 2010
NSF Site Visit to MIT DataSpace27
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace28
DataSpace
February 8, 2010
NSF Site Visit to MIT DataSpace29
DataSpace
Scientific Publication
• At present, all our rich interactive visualization tools are inside the lab
• When it’s time to publish, we use papers with dead ink charts
• Move to publication of rich visualization descriptions inside the article
• Show the reader not just conclusions, but the way you looked at your data to reach them
February 8, 2010
NSF Site Visit to MIT DataSpace30