Download presentation
Presentation is loading. Please wait.
Published byKerry Sharp Modified over 9 years ago
1
Service discovery with semantic alignment Alberto Fernández AT COST WG1 meeting, Cyprus, 15-16 Dec, 2009
2
Introduction Service coordination in open environments Identification of functionality (service) required Service provider discovery Service provider selection (trust, reputation, QoS) Service engagement (negotiation) Service invocation Agreement Technologies for Service coordination Semantics, negotiation, trust, …
3
Service Provider Discovery Matching service advertisements against service requests Service description languages Usually identical for advertisements and request
4
OWL-S service profile example INPUTS OUTPUTS CATEGORIES PARAMETERS EFFECTS/PRECONDITIONS Ontology
5
Service Provider Discovery Semantic Mismatches Service description models Domain ontologies
6
Architecture service description service request Model Alignment service description service request Service Matching Degree of match dom(C 1,C 2 ) Local Alignment KB Local Alignment KB Service Directory Service Directory Semantic Concept Matching Concept Alignment Registry Alignment Registry Different ontologies OWLS, WSMO, SAWSDL, WSDL, keywords, text,…
7
Service Model Alignment Service Description Approaches Semantic: OWL-S, WSMO Syntactic: WSDL Hybrid: SAWSDL Light models: keywords, tag-clouds, textual Common models for pairs of SD models Possible loss of expressiveness Integrated model for service discovery Union of common models
8
Integrated model for service matchmaking
9
ElementOWL-SWSMOSAWSDLWSDL Keyword (tag) Tag Cloud Text input xxxx--- output xxxx--- precondition xx----- effect xx----- keywords ---- x -- text xx----x category xxx- -- - tag cloud ----- x -
10
Integrated model for service matchmaking ElementOWL-SWSMOSAWSDLWSDL Keyword (tag) Tag Cloud Text input xxxx--- output xxxx--- precondition xx----- effect xx----- keywords O sem O syn x text xxO sem O syn x category xxxO syn -- - tag cloud O sem O syn x
11
Architecture service description service request Model Alignment service description service request Service Matching Degree of match dom(C 1,C 2 ) Local Alignment KB Local Alignment KB Service Directory Service Directory Semantic Concept Matching Concept Alignment Registry Alignment Registry
12
Service Matching Aggregation over matching of individual concepts (only for common fields) Current approaches to Semantic IOPE I A I R O R O A P A P R E R E A
13
Semantic Concept Matching Degree of Match between C A and C R. Combination of Level of Match subsumption relation Exact, plugin, subsumes, fail, … Concept Similarity Semantic distance
14
Semantic Concept Matching vehicle carvantruckbus Japanese car hondamazdanissantoyota American car chevydodgeford searched found plug-in levels of match
15
Semantic Concept Matching vehicle carvantruckbus Japanese car hondamazdanissantoyota American car chevydodgeford found searched subsumes levels of match exact > plug-in > subsumes > fail
16
Semantic Concept Matching vehicle carvantruckbus Japanese car hondamazdanissantoyota American car chevydodgeford distance = 2 concept similarity
17
Service Matching Non IOPEs (keywords, tag clouds, categories) Syntactic: Semantic:
18
Concept Alignment Alignments (or mappings) between two ontologies O and O’: where: e and e’ are the entities considered n: is a degree of trust (confidence) R is the relation holding between e and e’. Representation in RDF SPARQL for querying
19
Open Issues SPARQL as query language Two stage discovery process e.g. the requester doesn't know the inputs required Matchmaking completely in the directory? Private information Scalability Distributed directories Ontology alignments discovery
20
Conclusions Summary Architecture for service discovery Semantic alignment Common model for service descriptions Future work Implementation (currently) and evaluation Open issues pointed out
21
Service discovery with semantic alignment Alberto Fernández AT COST WG1 meeting, Cyprus, 15-16 Dec, 2009
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.