Title: III-SGER: Spatio-Temporal-Thematic Queries of Semantic Web Data: a Study of Expressivity and Efficiency
1III-SGER Spatio-Temporal-Thematic Queries of
Semantic Web Data a Study of Expressivity and
Efficiency
- Query Rewriting Framework for Spatial Data
- Funded by NSF AwardIIS-0842129 "III-SGER
Spatio-Temporal-Thematic Queries of Semantic Web
Data a Study of Expressivity and Efficiency - http//knoesis.org/research/semweb/projects/stt/
2Increased retrieval of spatial information
3But spatial querying still needs improvement
4User expected to pose proper query
5Whats the problem
- There is a need to analyze spatial information
and queries beyond lexical and syntactic level. - Datasets on which queries are executed might have
information, but not in the form expressed by
query constraints. - Question Find schools in NJ,
- Answer Sorry! Information is available for
schools in counties. - Semantic relationship such as part of between
query terms may or may not be understood properly
- Find Schools in Greene County
6What needs to be done?
- Information should be retrieved even if mismatch
exists between granularity of query constraints
and knowledge base - User should not be burdened with knowing how the
ontology is structured - Words which express semantic relations between
entities should be included in the query mechanism
7Proposed approach
- Ease writing of expressive queries to implicitly
exploit relations between entities. - Propose operators which hides the complexity of
user knowing and expressing named relations
between entities in query constraints - Behind the scene utilization and identification
of semantic relations such as partonomy in
answering queries. - Define transformation rules for operators to use
work by Winstons taxonomy of part-whole
relations. - Rule based approach allows applicability in
different domains such as life science with
appropriate modifications.
8Geonames schema modeling schools
County
School
Contains feature
Contains feature
Ohio
9SPARQL query for schema
- PREFIX rdf http//www.w3.org/1999/02/22-rdf-syn
tax-ns - PREFIX geo lthttp//www.geonames.org/ontologygt
- SELECT ?school
- WHERE
- ?state geofeatureClass geoA
- ?schools geofeatureClass geoS.
- ?state geoname "Ohio
- ?state geochildrenFeatures ?county .
- ?county geochildrenFeatures ?schools .
-
-
Users must know exactly how schools, counties,
and states are modeled, which is not trivial
10Our approach
SELECT ?school WHERE ?state
geofeatureClass geoA ?schools
geofeatureClass geoS. ?state
geoname "Ohio ?schools
in ?state .
Identify triple constraints to be replaced
Query Re-Writer
Triple constraints and variables rewritten using
transformation rules
SELECT ?school WHERE ?state
geofeatureClass geoA ?schools
geofeatureClass geoS. ?state
geoname "Ohio ?state
geochildrenFeatures ?county . ?county
geochildrenFeatures ?schools .
Rewritten constraints and variables plugged into
query envelope
Approach allows writing queries more naturally
and independent of the underlying knowledge base
11Architecture
- SELECT ?school
- WHERE ?school in Ohio.
User submits SPARQL Query
- Triple Constraints
- Query Variables
Mapping of ontology properties to Winstons
categories
Query Rewriting Engine
Domain Specific Transformation Rules
Meta rules for Winstons Categories
- Altered Triple Constraints
- Altered Query Variables
Rewritten Query according to the data structure
SELECT ?school WHERE ?state geoname
"Ohio ?state
geochildrenFeatures ?county . ?county
geochildrenFeatures ?schools .
12Evaluation plan
- Utilize publicly available datasets Geonames and
Sensor Ontology for identifying partonomical
relationships - Identify meaningful queries and their rewrite
- Expose infrastructure as a web resource for
utilization on other datasets.
13Publication/Community Contribution
- A Framework to Support Spatial, Temporal and
Thematic Analytics over Semantic Web Data
,Matthew Perry, Amit Sheth and Prateek Jain
(Under review in Transaction of Database Systems) - Submission for International Semantic Web
Conference in June 2009. - Sensor Ontology, linked with Geonames Dataset to
be made part of Linked Data Cloud
(http//linkeddata.org/data-sets).
14Thanks