ANCHOR-FLOOD: RESULTS FORANCHOR-FLOOD: RESULTS FOROAEI-2009
Md. Hanif Seddiqui and Masaki AonoKnowledge Data Engineering Laboratory
Toyohashi University of Technology JapanToyohashi University of Technology, Japan
OUR CONTENTS
Anchor-Flood for Ontology AlignmentBenchmarksBenchmarksAnatomyC fConferenceDirectory
I t M t hi Instance Matching IIMB Benchmarks
OUR CONTENTS
Anchor-Flood for Ontology AlignmentI t M t hi Instance Matching
O A A EONTOLOGY ALIGNMENT: AN EXAMPLE
pig
Ontology BOntology A Aligning TWOOntologies ?
animal
pig
mouseNow with Anchors?
Ontologies ?NxM comparisonsIn Brute-Force
fish
OUR ONTOLOGY ALIGNMENT TECHNIQUE
Ontology AOntology B
pig
Aligned
Anchor-Flood works faster !!
ANCHOR-FLOOD ALGORITHM
FINDINGSFINDINGS
Block Size vs Elapsed TimeBlock Size vs. Elapsed TimeTwo depth children from anchor-concept c+ one depth children from parents(c) +
Grandparents/ 1 depth
p p ( )one depth children from grandparents(c)on anatomy track
Parents / 1 depth
Elapsed time approx. 15 sec.Decreasing recallGood precision
Anchor-concept / 2
depthGood precision
Two depth children from anchor-concept Grandparents/ 1 d th
depth
Two depth children from anchor concept c+ two depth children from parents(c) + one depth children from grandparents(c)
1 depth
Parents / 2 on anatomy trackElapsed time approx. 4 min.I ll
Parents / 2 depth
Increase recallDecrease precision
Anchor-concept / 2
depth
OUR CONTENTS
Anchor-Flood for Ontology AlignmentI t M t hi Instance Matching
SEMANTIC LINK CLOUD: OUR UNIQUENESSSEMANTIC LINK CLOUD: OUR UNIQUENESS
Resource Audio Album
depicts
Time
CreatorMultimediaContentTitle
Nov, 2007
hasCreatorhasCreationDate
Creation
Britney Spears
BlackouthasTitle
Blackout
Britney Spears
USA
hasCreationLocation
an instance is defined as a part of knowledge that includes li k d t ti th i tlinked concepts, properties, other instances
to specify its meaning. We call this as a ‘Semantic Link Cloud’
INSTANCE MATCHING ALGORITHM
INSTANCE MATCHER
RESULTS
Please visit OAEI-2009 website for the detail results of aflood stands for Anchor Floodresults of aflood, stands for Anchor-Flood
CONCLUSIONS AND FUTURE WORK
Anchor-Flood algorithm run faster due to its unique way of divide and conquer F I M hi d S i li k i d For Instance Matching, we used Semantic links associated to each of the Instances.
Future WorkFuture WorkTo consider Semantic Similarity among concepts of a taxonomy to reduce the size of block and hence to decrease th ti d t i th ffi ithe runtime and to increase the efficiency.Improve the runtime of Instance Matching
You can download our system through-www.kde.ics.tut.ac.jp/~hanif/res/2009/anchor flood.zipwww.kde.ics.tut.ac.jp/ hanif/res/2009/anchor_flood.zip
Related Paper: Md.H. Seddiqui and M. Aono, An Efficient and Scalable Algorithm for Segmented Alignment of Ontologies of Arbitrary Size, Web Semantics (to be published)( p )
TTHANK YOU
CHALLENGES
Varying Block SizeIncrease block size by the neighbors of sufficient depthIncrease block size by the neighbors of sufficient depthDecrease block size by considering semantic similarity
Varying threshold Varying threshold