© 2003 Eindhoven University of Technology Alexandru Telea, Flavius Frasincar, Geert-Jan Houben...

Post on 19-Dec-2015

214 views 1 download

Tags:

Transcript of © 2003 Eindhoven University of Technology Alexandru Telea, Flavius Frasincar, Geert-Jan Houben...

© 2003 Eindhoven University of Technology

Alexandru Telea, Flavius Frasincar,

Geert-Jan HoubenEindhoven University of Technology, the Netherlands

Visualizing RDF(s)-based Information

• What is RFD(s) data• Visualizing RDF(s) data• The Gviz tool• Applications• Conclusions

Overview

What is RDF(s) data?

RDF: resource description framework (http://www.w3.org)

• two graphs: instance and schema

• foundation for exchanging metadata• describes web resources

named resource URIanonymous resource -literal name

Node Type Value

property URI

Edge Type Value

Questions and Requirements

• understand and modify the RDF(s) data

• understand: RDF(s) data = graphs graph understanding visualization

• modify: graph editing

• typical questions: how is a RDF(s) dataset looking? does an instance match a schema? how does an instance evolve? how does a schema evolve?

need for visual graph analysis/editing tools

Previous Work

Text-based tools: Protégé-2000

‘Newspaper’ example: list of articles, sections, employees, advertising in a fictitious newspaper.

• text-only

• insightful only for small datasets

• not easy to add ‘what-if’ queries and scenarios

Previous Work

Visual tools: RDFSViz visualization tool

Uses AT&T GraphViz’s graph drawing to display RDF dataLimited to directed DAG drawing layouts

• visual

• insightful only for small datasets

• not easy to add ‘what-if’ queries and scenarios

Previous Work

Visual tools: OntoViz plugin for Protégé

Enhances Protégé with graph drawing capabilitiesSame (limited) directed layout as RDFSViz

Previous Work

Visual tools: IsaViz

• visual

• set of graph editing tools

• insightful only for small datasets

• not easy to add ‘what-if’

queries and scenarios

Goal

Provide a visual examination and editing tool forRDF(s) data that:

• copes with realistically large datasets• allows an easy definition of new queries• allows an easy definition of new visualizations (layouts, coloring schemes and shapes, etc)

Can we reuse/adapt an existing tool?

The GViz Tool

• first used in the context of reverse engineering (thus handles large graphs) (VisSym’02, IWPC’02, TOOLSEE ’02)

• generic data and operation model

• allows end-user customization of all operations: - selection: what to display - layout: how to arrange - glyphs: what to draw - interaction: how to respond

GViz Architecture Overview

RDF(s) data

selected data

displayed data

input

query

display &interaction

GViz Operation Pipeline

layout 1 (GEM) layout 2 (dot) type-colored glyphs

initial data

selected subset selected subset 2

... other operations …

Newspaper Example - Comparison

IsaViz GViz

yellow: literalsgreen: resources

red: subclassOfblue: typewhite: others

Nodes Edges

orange: nodes with a Property edge

Applications

• customizable selections• schema-instance comparisons - how/what of a schema is instanced• instances comparison - how do two instances (of same schema) differ• schemas comparison - how do two schemas differ (e.g. schema evolution)

graph comparison operations(done only for non-anonymous nodes)

Applications

RDF(s) work data:

User Agent Profiles (UAProf) = RDF(s) datasetsdescribing mobile phone capabilities

Example:

UAProfschema

literals

resources

nodes with aProperty edge(towards literals)

subclassOfedges

Customizable selections

full schema only edges from/toclicked component

user clicks this component

customizing selection script: 18 Tcl linescustomizing glyphs script: 40 Tcl lines

Schema-instance comparisonschemaNokia 8310 instance instance vs schema

Use shape for type: named literals anonymousUse color for comparison: instance schema common

Most instance-specific nodes are literals (yellow, )Only the (few) component-types are instantiated (red, )Many uninstantiated properties (green)

Instance-instance comparison

instance specificEricsson onlyall four phones

Color usage

similar overall structure

specific: literals, e.g.phone name, etc.Only one commonresource found!This led to discoveringan inconsistent namingscheme betweendatasets

Two Ericssons moresimilar than rest

Schema-schema comparison

schema specific2000, 2001 only2001, 2002 onlyall years

Color usage

little gray in (2000,2001), soschemas are very similar

no yellow!!! so nothing onlyin 2001 and 2002enough red in (2001,2002), sopart common to all years kept

2000 2001

2001 2002

2002 is a new product familywhich breaks the 2000-2001continuity

Conclusions

Combination of customizable selections, glyphs,layouts, and interaction is very effective for understanding RDF(s) datasets

Facts found by visualization (and previously unknown):• naming scheme changes• mobile phone instances for different schemas are similar• product family breakpoint in schema evolution

Effort needed to adapt Gviz tool to RDF(s) data & tasks:• 10-40 Tcl lines per task• 30 minutes for the first task, 5-10 minutes afterwardsNo need to develop new tool

Conclusions

A flexible graph visualization tool allowing easy end-usercustomization of most operations is essential

Spring-embedder layouts more effective than directedtrees/DAGs if combined with selection and glyphs

Need to look at• RDF(s) data editing• metrics for selection and glyph parametrization

Appendix: Mapping and Visualization

Map ‘abstract’ graph data to ‘concrete’ visual form

Mapping and visualization pipeline

Appendix: Mapping and Visualization

Basic Mapping

mappers data->2D/3D geometries

viewers geometries->display

glyphs parameters->geometries

glyph factories attributes->parameters

graphdata

mapper

Glyph factory

glyphs viewer

Appendix: Mapping and Visualization

Glyphs• similar to the SciVis glyphs• 2D/3D parametrizable graphical objects• implemented as (small) Inventor scene graphs

Glyph Factories• called by mappers for each node/edge to map• written as (small) Tcl scripts, thus very easy to customize• selectable/editable at run-time to map data in various ways

Appendix: Mapping and Visualization

Appendix: Mapping and Visualization

Advantages of the chosen architecture:

• easy to produce different mappings on the fly (average Tcl glyph factory < 15 lines of code)

• flexible (control mapping at node/edge level)

• simple to implement (2 mappers vs >20 in SciViz)

• adding more complex mappers could e.g. produce UML-like diagrams automatically