A DH Tbased A pproach for R esource M apping through A pproximation Luca Maria Aiello Marco Milanesio Giancarlo Ruffo Rossano Schifanella Università degli Studi ID: 463175
Download Presentation The PPT/PDF document "Tagging with DHARMA" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Tagging with DHARMAA DHT-based Approach for Resource Mapping through Approximation
Luca Maria Aiello, Marco MilanesioGiancarlo Ruffo, Rossano SchifanellaUniversità degli Studi di TorinoComputer Science Department
Keywords : navigational search, folksonomy, DHT, approximated graph mapping, last.fm
Seventh International Workshop on Hot Topics in Peer-to-Peer Systems
Speaker: Luca Maria Aiello, PhD student
aiello@di.unito.itSlide2
OverviewGoal: enrich the p2p layer with a tag-based navigational search engineTask: mapping a folksonomy on a DHTProblem: mapping dense graphs on distributed layer is very inefficientSolution: approximated mapping using a complexity-bounded algorithmEvaluation: the approximated solution does not upset the structure/semantic of the folksonomy23/04/2010
HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino2Slide3
MotivationsDirect search Navigational searchTaxonomies (e.g. Yahoo! directory): not successfulFolksonomies: successfully emerging thanks to the social webTag-based search engine on DHTs is profitableApplications layered on DHTs use exact match search (e.g. eMule) folksonomic search adds flexibilityRecent research activity on P2P for privacy-aware online social networksFolksonomic
search engine is an important OSN feature23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino3Slide4
Folksonomies structureRepresented as a tripartite hypergraph23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino4
MetallicaIron Maidenmetalclassic
JohnSlide5
Tag-Resource Graph (TRG)Projection on user dimension23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino5Metallica
Iron Maidenmetalclassic
John
Bipartite graph
Edges are weighted depending on number of tag assignments
3
5
1Slide6
Folksonomy graph (FG)Models tag-to-tag similarity with co-occurrence23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino6
MetallicaIron Maidenmetalclassic
3
5
1
2
5+3=8
1+2=3
8
3
Co-occurrence similarity
is widely known and used
Local
calculation
DirectionalSlide7
Folksonomic searchUser selects a tagRelated resources and tags are displayedRanking based on arc weightsUser can shift to another tag or select a resource23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino7
metalthrashpower
grind
classic
Iron Maiden
Metallica
Manowar
8
3
12
1
5
7
9Slide8
Folksonomy evolutionThe folskonomy grows quickly due to massive user activityResource insertionTag insertion23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino
8Both FG and TRG should be updated properly!Slide9
Folksonomy maintenance23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino9Slide10
Mapping on a DHTMap FG and TRG on the DHT and implement the update operations on the distributed layerIdea:Splitting the FG and TRG into small structural parts or modulesEach module contains a node (tag or resource) and its outgoing edges/arcsEach module is mapped on a different p2p index node23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino10Slide11
Mapping FG on a DHT23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino11Slide12
Mapping TRG on a DHT23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino12Slide13
Folksonomy maintenance on the DHTWhen inserting a new tag or resource, proper modules have to be modifiedComplexity of resource insertion and search are linear with the input size OK!Tagging operation is linear with |Tags(r)|Tags(r) is the set of tags labeling the resource rHow many tags for a resource?Let’s see a real-world example…
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino13Insert(r, t1,…, tm)Tag(t,r)Search step#lookups2 + 2m
4 +|Tags(r)|2Slide14
Last.fm folksonomy sample99,405 users, 1,413,657 items and 285,182 different tags23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino14
|Tags(r)|Can be huge!Slide15
An approximate solutionDHARMADHT-based Approach for Resource Mapping through ApproximationIdea:Put a constant upper bound to the number of lookups for tagging operationThe resulting FG will be approximated…
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino15Slide16
Approximate tagging: k=1
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino16metal
thrash
classic
cool
Lookup count :
0
Metallica
1
3
4
Now, select
k
tags at random in Tags(Metallica) and link them with the new tag
5Slide17
Approximate graph vs real graphWith DHARMA approximation, the complexity is affordable for very small kThe smaller k is, the more the structure of the Folksonomy Graph is upset! We need to compare the approximate and the real Folksonomy Graph23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino
17Insert(r, t1,…, tm)Tag(t,r)Search step#lookups2 + 2m4 + k
2Slide18
Approximation in Last.fmWe simulate the evolution of FG with the approximated protocolSimulated resource insertion and tagging activityGoal: draw a comparison between the real Last.fm FG and the approximated FG, for different values of kNodal degreeArcs weight23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino
18Slide19
Nodal degree comparison23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino19Slide20
Arc weights comparison23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino20Slide21
Keeping proportionsSimulated and real graph are different but……they should not necessarily be equalOnly proportions must be keptArcs weight ordering must be keptProportion between weights is not lost (no flattening)Only the “less informative” arcs should be deletedSame proportions grant the preservation of the navigational semantic
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino21Slide22
Keeping proportions: k = 1Arcs weight ordering is keptKendall’s tau measure between arcs weight rankings is very high (0.7) OK!Proportion between weights is not lost (no flattening)Cosine similarity between arcs weight sets is very high (0.8) OK!Only “less informative” arcs are deleted40% of arcs gets lost, but 99% of them has weight <= 3
OK! (we wipe out only noisy links)23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino22Slide23
Query convergence rapidity: resultsStepsLastRandFirstOriginalgraph
μ3.476.41233.94σ1.41754.458715.9942
μ1/235
33
Approximatedgraph
μ
3.38
5.2140
19.17
σ
1.2373
2.6994
10.3065
μ
1/2
3
5
16
23/04/2010
HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino
23
A query converges
through subsequent filtering on the resource set
Estimation of convergence with simulation
Convergence is
quick
Simulated search has no semantic
notion
Approximation
speeds up the searchSlide24
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino24Slide25
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino25Slide26
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino26Slide27
23/04/2010HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino27Slide28
ConclusionsApproximated folksonomy mapping allows an efficient P2P implementationThe information lost in approximation is prevalently noise (automatic filtering)Search navigation realizes vocabulary specialization, converging to narrow semantic categoriesConvergence is quick (and even quicker with approximation)DHARMA (Java implementation) is available at: http://likir.di.unito.it/applications23/04/2010
HOTP2P 2010 - Luca Maria Aiello, Università degli Studi di Torino28Slide29
Related worksOther attempts in mapping folksonomies on p2p systems:E.g. Tagster [1]Privacy-aware P2P online social networksSafebook [2], PeerSon[3], Likir[4,5]29/03/2010SESOC 2010 - Luca Maria Aiello, Università
degli Studi di Torino29[1] Görlitz, Sizov, Staab – ESWC 2008[2] Cutillo, Molva, Strufe – WONS 2009[3] Buchegger, Schöiberg, Vu, Datta – SocialNets 2009[4] Aiello, Milanesio,
Ruffo, Schifanella – P2P 2008[5] Aiello, Ruffo – SESOC 2010Slide30
Speaker: Luca Maria Aiello, PhD studentaiello@di.unito.itThank you for your attention!
Seventh International Workshop on Hot Topics in Peer-to-Peer Systems