Title: DLDB2: A Scalable MultiPerspective Semantic Web Repository
1DLDB2 A Scalable Multi-Perspective Semantic Web
Repository
- Zhengxiang Pan, Xingjian Zhang and Jeff Heflin
Presented on Web Intelligence (WI08) December
11th, 2008
2Overview
- We present a Semantic Web Knowledge Base System
that scales both in terms of number of ontologies
and quantity of data. It also supports reasoning
from different view points. - By delegating TBox reasoning to a DL reasoner, we
focus on the design of the table schema, database
views, and algorithms that achieve essential ABox
reasoning over an RDBMS. - We evaluate the performance using benchmarks and
real world data from multiple sources.
3Ontology perspectives
- There may not be a universal model of all the
data sources and ontologies on the web - Extremely unlikely that one could exist (e.g.
inconsistency) - Instead, we should allow different viewpoints and
contexts, which are supported by different
ontologies - Each perspective is based on an ontology
- An ontology provides a shared context
- Data sources that commit to the same ontology
implicitly agreed to share a context - When it makes sense, data that commit to
different ontologies can be included to maximize
integration
4Ontology perspectives -cont
- Theoretically, each perspective represents a
certain view of the world, and could be
considered a knowledge base. - The answer to a semantic web query must be
relative to a specific perspective - A mapping ontology between a sequence of
ontologies Oi ,..., On introduces no vocabulary
of its own but extends Oi ,..., On and contains
the axioms that map their vocabularies.
5Integration using perspectives
Car
Automobile
O1
O2
commits to
commits to
ce7334?O1Automobile
imports
ezz3290?O1Car
O1Car(x) ? O2Automobile(x)
Om12
O1Car(x) ? O2Automobile(x)
Om12
?ltO1 , car(x)gt ? ezz3290 ?ltO2 , automobile(x)gt
? ce7334 ?ltOm12 , car(x)gt ? ezz3290,
ce7334 ?ltOm12 , automobile(x)gt ? ezz3290,
ce7334
?ltOm12 , car(x)gt ? ezz3290 ?ltOm12 ,
automobile(x)gt ? ezz3290, ce7334
6DLDB2 System
- A knowledge base system
- extends relational databases with partial OWL
reasoning capabilities - Uses DL reasoner to precompute subsumptions
- Architecture
- The components are loosely coupled
- Any DIG compliant reasoner
- Any SQL compliant RDBMS with JDBC driver
- Reasoning services and databases can be
distributed on different machines or clustered on
multiple machines
7Table Design
schema-aware and decompositional
1Student
Ontologies_Index
SeqNum
URL
SeqNum
URL
1
http//www.lehigh.edu/zhp2/univ
-
bench.owl
1
http//www.lehigh.edu/zhp2/univ
-
bench.owl
Source_Index
SeqNum
URL
SeqNum
URL
1
http//www.lehigh.edu/zhp2/univ
-
bench.owl
1
http//www.lehigh.edu/zhp2/univ
-
bench.owl
2
http//www.lehigh.edu/ubdata/univ0_0.owl
2
http//www.lehigh.edu/ubdata/univ0_0.owl
1TakesCourse
URI_Index
ID
URI
ID
URI
1
http//www.Dept0.Univ0.edu/UndergraduateStudent121
1
http//www.Dept0.Univ0.edu/UndergraduateStudent121
2
http//www.Dept0.Univ0.edu/GraduateCourse9
2
http//www.Dept0.Univ0.edu/GraduateCourse9
3
http//www.Dept0.Univ0.edu/GraduateStudent123
3
http//www.Dept0.Univ0.edu/GraduateStudent123
8Inference
Load ontology
Load data
Data
Ontology
Student ? Person who takes a Course GraduateStud
ent ? Person who takes a GraduateCourse GraduateCo
urse ? Course
Find individual equivalence
DL Reasoner
Graduate Student ? Student
resolve individual equivalence
Inferred Hierarchy
Compute Transitive Closure
CREATE VIEW 1Student_1_view AS SELECT FROM
1Student where onto1 UNION SELECT FROM
1UndergraduateStudent_1_view UNION SELECT FROM
1GraduateStudent_1_view
table view creation
Store in database tables
Database operation
sequences numbers of the ontologies being both
ancestors of current ontology O and descendants
of the ontology that defines the term
9Query Answering
Query Interface application
http//swat.cse.lehigh.edu/ubmap.owl SELECT
?XWHERE ?X rdftype GraduateStudent. ?X
takesCourse lthttp//foo.edu/course0gt.
SPARQL query
Query API
SELECT 1GraduateStudent_2_view.ID FROM
1GraduateStudent_2_view, 1takesCourse_2_view
WHERE 1GraduateStudent_2_view.id
1takesCourse_2_view.subject AND
1takesCourse_2_view.object http//www.foo.edu/d
ept0/course0
Query Translation Algorithm
SQL Sentences
RDBMS
10Benchmark Load Time
11Benchmark Query Time
12Benchmark -Completeness
Completeness on UOBM Lite 5
- DLDB2 is complete on all LUBM queries
- DLDB2 on UOBM
- complete on 10 out of 13 queries
- 2 of the incomplete queries are due to universal
restrictions. The rest is due to cardinalities - DLDB2 is complete on a sizable subset of
Description Horn Logic - plus more DL inferences on TBox
13Multi-ontology Evaluation -Data
Loaded data sources
14Multi-ontology Evaluation -Load
15Multi-ontology Evaluation -Query
- In our evaluation, we first create the mappings
between pairs of related ontologies, and then map
them to popular ontologies like FOAFand Dublin
Core. - Totally, we created 16 mappings with 46 axioms.
We also created 4 common perspectives such as
academic and e-government each perspective
extends (imports) some mappings of interest. - Additionally, we created 169,023 owlsameAs
statements to map URIs from different sources but
refer to the same individual.
Query performance
16Future work
- Support efficient updates on documents
- Keep the data fresh
- Improve query performance
- Utilize query optimization techniques
- Prove the completeness of the system
- Formally define the language that the system is
complete on
17Thank you