UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic...

13
UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej Kao, Kerstin Voss [email protected] , [email protected], [email protected] , [email protected]

Transcript of UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic...

Page 1: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

PeerThingP2P-based Semantic Resource Discovery

Felix Heine, Matthias Hovestadt, Odej Kao, Kerstin [email protected], [email protected], [email protected], [email protected]

Page 2: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Agenda

Motivation

System Design and Workflow

Implementation Details

Conclusion

Page 3: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Motivation

Search:

Pentium 3

RAM >= 256 MBPentium 4

SRAM = 1024 MB

Pentium 3

DDR2 RAM = 512 MB

Organization AOrganization B

Organization C

SemanticSearch Engine

Page 4: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Motivation

Previous work for semantic search Central database stores information DL (Description Logics) more expressive than RDF

(Resource Description Framework)

Decentralized concept necessary Local resources should be favored Search according to static and dynamic data

Page 5: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Idea

Page 6: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Pool

Poolclient Poolclient

System Design

Userclient

Poolhead Pool

Poolhead

Poolclient

P2P

Poolclient

Poolclient

Page 7: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

PeerThing - Initialization

Pool

Poolhead Pool

Poolhead

Poolclient

P2P

Poolclient

Poolclient

Poolclient Poolclient

S

S

S S

S

S

S

Userclient

Page 8: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

PeerThing – Query Process

Pool

Userclient

Poolhead Pool

Poolhead

Poolclient

P2P

Poolclient

Poolclient

Poolclient Poolclient

S

S DS D

D

S DD

Pentium 3500 MB RAM free

P3, 250 MB free

P3, 600 MB free

P2, 500 MB free

S

AMD XP AMD Athlon

Page 9: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Local Implementation

Poolclients Elect one poolhead Inform poolhead once about static features Provide dynamic data on request

Poolheads Expert system RACER PRO for semantic search

Knowledge base in OWL DL (Web Ontology Language) Requests in DIG

Hold static data from his clients, cache dynamic data Local resources can be favored in search Establish connection to other pools via P2P

Page 10: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Features of the P2P Network

Bases on Gnutella concept (JTella implementation) Modifications:

Locality is considered for the connection process Connect only with the “nearest” peers

Direct Routeback of answers

Page 11: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Features for Usability

Semantic Queries No detailed description necessary Compatible resources can be found

Higher success rate, easy to use

UserInterface shows currently used ontology Dynamic reconfiguration Only valid requests Language reconfiguration according to ontology labels

Easy to use, to reconfigurate

Page 12: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Screenshot

static dynamic

Page 13: UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft PeerThing P2P-based Semantic Resource Discovery Felix Heine, Matthias Hovestadt, Odej.

PeerThing: P2P-based Semantic Resource Discovery

UNIVERSITÄT PADERBORN Die Universität der Informationsgesellschaft

Conclusion

Semantic search finds compatible results Decentralized concept with no bottlenecks Dynamic reconfiguration with ontologies Simple usability Improved P2P Network implementation Prototype implemented

Future Work: Support of multiple ontologies

Thanks for your attention!Thanks for your attention!Questions?Questions?