Transcript Partial Satisfaction Planning (PSP)

GEONSearch: From Searching to
Recommending
Ullas Nambiar, Bertram
Ludaescher
Ghulam Memon, Dogan Seber,
Chaitan Baru
Dept. of Computer Science
University of California, Davis
San Diego Supercomputer Center
University of California, San Diego
GeoInformatics 2006 May 10-12, Reston, Virginia
 NSF Large ITR project – collaborative effort
 GEON is creating an IT infrastructure to “enable”
interdisciplinary geoscience research -- not a group of
researchers, but the entire community will benefit
 Support efficient Knowledge Discovery from GeoScientific
Data
– GEONPortal provided as a Web-based tool for knowledge discovery
Scientific Knowledge Discovery
GEON Map/Query
Integration & Tools
GEONSearch
GEON Network
Source: Fayyad, U., Piatetsky-Shapiro, G., Smyth, P., Uthurusamy, R., Advances in Knowledge Discovery and Data Mining, 1997.
Prototypical query over GEON
 Query: “Find Gravity data for regions near Rocky
Mountains where geologic age is Jurassic”
 What is needed ?
– Definition of “regions near” Rocky Mountains
• Approximate “regions near” as “Rocky Mountain States”
 Colorado, Idaho, Montana, Nevada, Utah and Wyoming
– Geologic age data for above regions
– Gravity measurements taken in those regions
– Map of continental US or that of Rocky Mountain States
 How to obtain it ?
– From GEON and with the help of GEONSearch
Current GEONSearch Features
 GEONSearch is the resource discovery tool available under GEON
Portal
 Allows users to retrieve datasets by querying its
– Keywords, Title and Description
– Metadata
• Search the Subject Taxonomy under which datasets are classified
– Spatial Coverage
• A bounding box of region covered by a registered dataset
– Temporal Coverage
• Age of objects evaluated or time when evaluations done
– Concepts from Ontology
• Concepts to which dataset or items contained in dataset are mapped to during
registration
Keyword Search in GEON
Rockies Jurassic or Colorado
Jurassic or Jurassic ……
Rockies gravity data or Colorado
gravity or Montana gravity or
gravity data ….
Rocky Mountain Region Map or
Colorado Map or state maps….
Too many answers
Too few answers
Advanced GEONSearch
Enter the potential subject group
of datasets ?
Datasets containing a city in
Colorado, Idaho, etc
Data with age Jurassic
Datasets mapped to concept mountains or
gravity or geologic age
Improving GEONSearch
 The challenge before GEONSearch
– Reduce iterative querying effort from Scientists
 The solution
– Suggest “similar queries”
• Queries with more/less keywords
• Queries that likely to have highly similar answers
– Suggest “related answers” for every result based on
•
•
•
•
Spatial proximity
Temporal proximity
Usage patterns
Common Ontology mappings
 Caveat
– Suggested queries and answers must be ranked using corresponding
distance measures
Knowledge Discovery using GEON
Need Gravity data
around Rocky
Mountains ……
Integration Cart (MyGEON)
RD1V1
GEONSearch
DD2V4
Query/Map
Integration
DD7V1
GEON Network
RD1V1
RD2V2
DD6V1
DD2V4
DD2V2
Registered Datasets
Derived Datasets
Ontology
Tools
M
e
t
a
d
a
t
a
How to share
result with other
scientists !
DD6V1
Register
DD7V1
Supporting Versions in GEON
 GEON network can be visualized as a virtual scientific database
that stores results of scientific inquiry
– Every dataset available reflects a scientific process
• E.g. Dataset D1 contains gravity measurements around Davis
 Datasets may change over time
– Parameters of a scientific process can be changed resulting in additional
results
– Updating a registered dataset could affect outcome of other scientific
inquiries
• E.g. Updating D1 with new data may make results of processes using D1
irreproducible
GEON Versioning System
 Focus
– Revision management for locally-hosted data
• ASCII, Excel, ESRI Shapefiles, GeoTIFF, OWL files etc
 Operations supported
– Revisions
– Branching
 Assumptions
– Dataset provider decides between Branch or Revision
– No support for eventual merging
Provenance Management
 Provenance Management becomes necessary in presence
of versioning
 The provenance of a piece of data is data about process
involved in generating the data.
– Who, What, Where, How, Assumptions
a2
a1
a2
transformation
a1
a1 a2
 Useful for verifying quality of datasets recommended by
GEONSearch
Summary
 GEONSearch is a necessary tool for knowledge discovery
using GEON
– Allows both simple Keyword Search and advanced Search
 GEONSearch will soon be enhanced to provide “related
content” and thus improve the “Search Process”
 GEON Versioning System under development for
supporting version management and provenance tracking
Thank You!
Feedback is very welcome:
 Questions
 Suggestions
 Specific Use Cases
Contact Info:
Search for “Ullas Nambiar” in your favorite Search Engine