OWLVisual Programming

1
OwlViperWeb Ontology Language Visual
Progamming for Data Reduction (of work)

• Ed Shaya,
• Brian Thomas,
• Zhenping Huang,
• Peter Teuben
• (Astronomy Dept. at Univ. of Maryland, USA)

2
Next Generation Analysis Software
Layer on top of VO infrastructure with
Scientists graphical interface. Environment in
which the scientists asks scientific questions or
states their goal.

• Ingredients OWL, Registries, Grid, WebServices,
  WSDL, OWL-S, Java, Math Packages, existing
  scientific analysis software.

3
Reasoners

• Facts, Jesse, Jena, etc can reason from Classes
  and Instances to determine self-consistency of an
  ontology (Classes with properties and comlex
  relationships between these) and deduce implicit
  classes and classify instances.
• CLIPS and other rules analyzers can start with
  the rules of a game (or problem) and determine a
  (usually optimal) strategy.
• Combining these software with hardware of future
  (supercomputers are now at the computational
  speed of a mouse) imply human level logical
  reasoning in a decade.
• Computers only lack the knowledge base.

4
H0 Determination on OWLViper Canvas
SpiralGalaxy Distance Redshift
5
Functionality

• Visual Programming
• Query the VO distributed data centers, analyze,
  visualize, and web-publish results.
• UML concept draw flow diagram, press button for
  code generation.
• A layer on top of normal code, pipelined.
• Various levels of autonomy
• Lowest level manual creation of a flow diagram
• Highest level state the goal and let it rip
  via OWL knowledge base.
• Scientific - it must handle unit conversions and
  error propagation as well.

6
Ontology (OWL)

• Functions and data are placed in class structure
  (not a strict hierarchy) plus properties or
  relationships.

Code
relation
Class1
output
input
Operation
inverse
7
Teach your computer well (with OWL)

• Create Subclass
• Mass SubClassOf Measurement
• hasUnits restricted to MassUnit
• hasValue
• hasError
• EarthMass InstanceOf MassUnit.
• consistsOf uikilogram
• factor 5.9742e24
• MassOfEarth InstanceOf Mass
• hasUnits EarthMass
• hasValue 1.0
• hasError 2e-5
• of Earth
• Ultimately create an ontology of 5000 terms and
  relationships between them. All fits in RAM.

8
Classes and Instances in Query

• The data in repository should be logically seen
  as instances of established Classes restricted by
  the constraints of the observation and with a
  subset of appropriate properties available.
• A query for all objects satisfying a set of
  constrains should be framed as a Class, ie a
  Subclass of an established class with user
  restrictions on the range of property values.
• Queries for data on a particular object are
  simple RDQL
• Mars hasPart ?S ?S isa Sky ?S hasColor ?C
• ?C of ?S ?C isa Color ?S isa Sky ?S isPartOf
  Mars
• The Query can be distributed to relevant
  datacenters where a reasoner determines all local
  instances of query class.
• We are writing a translator from Owl to ADQL.

9
Table_row.owl

• astSpiralGalaxy
• sciPositionAtTime
• scihasLocation
• astEquatorialCoordinatesB1950
  rdfIDtable1475
• geohasCoordinates
• astRA_B1950
• qhasValue qValue
  rdfIDRA_col
• qhasAccuracy
• statError qhasValue
• qValue rdfIDRA_err_col
  
• astDEC_B1950
• qhasValue qValue
  rdfIDDE_col
• qhasAccuracy
• statError qhasValue
• qValue rdfIDDEC_err_col
  
• scihasReference sciReference
  rdfIDref23
• sciat sciForever

10
Find an AGN that has detection in radio-band and
is in the Virgo Cluster and has luminosity gt 3e9
SolarLuminosity and has spectrum in UV.

• G1
• subclassOf
• AGN
• Name?name
• hasDetection  
• Detection inEMBand RadioBand         
• hasLocation 
• Direction in VirgoClusterRegion
• (Instance of Polygon on Celestial Sphere)
• EquatorialJ2000Coordinates?J2000
• hasMeasurement
• BolometricLuminosity hasValue
• minValue 3e9
• HasUnits SolarLuminosity
• hasSpectrum           
• Spectrum
• inEMBand
• UV
• ?spectrum

11
(No Transcript)
12
(No Transcript)
13
Conclusionshttp//archive.astro.umd.edu/ont/index
.html

• Complex scientific query plus subsequent
  transforms via GUI or pseudo-natural language is
  seems feasible, but requires development of
  ontology, catalog data to have owl profile, and
  simple reasoners at data centers.
• Application to find and analyze data must have
  ability to
• Find operations by name or owlClass of
  output/inputs data.
• Starting from goal (usually a subclass of
  existing classes), analyzer finds path back to
  source data.
• Time estimator
• Logic evaluator
• Basic mathematics at main level
• Error propagation
• Unit analysis, conversion
• Tests for infinite loops
• Interfaces C, Java, WebServices, IDL, Matlab,
  pyRAF, Ftools, Numerical Recipes