Evgeny Zolin, School of Computer Science, University of Manchester, UK, [email protected]
Andrey Bovykin, Department of Computer Science, University of Liverpool, UK, [email protected]
Abstract
Examples of services and matches
We present a formalism for describing Semantic Web Services
A service request Q1: a user is looking for a service that
returns the list of wines that are sold in a specified region:
Main features of the approach:
 It deals with information providing (stateless) services
 Enables for service discovery of high precision / recall
 Service descriptions use terms from background ontology
 The problem of matching web services is decidable
 Compatible with standard approaches (OWL-S, WSMO)
y1
x2
xm
z2
zr
yn
Outputs
Inputs
z1
Note: the services S and Q1 have the same inputs/outputs,
but they perform different functions, so they do not match.
A service request Q2: a user is looking for a service that
Describing Services
x1
Input:
g: GeoRegion
Output: w: Wine
Relationship: THERE IS SOME s:
( s: Shop & s LocatedIn g & s Sells w )
returns the french wines produced in a given french region:
Input:
g: FrenchGeoRegion
Output: w: FrenchWine
Relationship: THERE IS SOME f:
f: WineGrower & f LocatedIn g & f Produces w
Note: here, the service S matches Q2, but in “standard”
approaches it (mistakenly) does not, since the outputs of S
(Wines) are broader than the user desires (FrenchWines).
A Service Description consists of:
specification of inputs xi and
their types Xi (the information
accepted by a service)
specification of outputs yj and
their types Yj (the information
returned by a service)
specification of relationships between the inputs and outputs
(which has the form of a conjunctive query, see [2] for details)
Example of a service description
A service advertisement S: a service returns the list of
wines that are produced in a specified geographical region:
Input:
g: GeoRegion
Output: w: Wine
Relationship: THERE IS SOME f:
f: WineGrower & f LocatedIn g & f Produces w
Note: terms “GeoRegion”, “LocatedIn” etc. come from a
geo-ontology; “Wine”, “Produces” etc. from a wine ontology
Service Matching
A service advert S matches a service request Q, where:
S has input type X, output type Y, relationship F(x,y)
Q has input type Z, output type W, relationship G(x,y)
w.r.t. a background ontology T if two conditions hold:
Applicability: Z is a subtype of X w.r.t. T ( Z
X )
Coherence: the services S and Q
return the same
answers on any input a from Z: for any individual b, the
conditions W(b) and G(a,b) hold iff Y(b) and F(a,b) hold.
always1
_____________________
1 For any data (i.e., an ABox) that conforms the scheme T. See [2] for detailed definition.
Services
repository
Service
request
Ontology
DL Reasoner
Service matchmaker
Results:
Service 1
Service 2
…………
Comparison to related approaches
• In the presented approach, service matching problem is
decidable. But only info-providing services are covered.
• In OWL-S [1], a Service Profile describes IOPE (inputs,
outputs, preconditions, and effects), thus stateful services
are covered. But for stateless services, it has no way of
relating inputs and outputs, due to limitations in the OWL.
• The Web Services Modelling Ontology (WSMO) has a
mechanism for relating inputs to outputs, but the resulting
matching condition is not decidable, due to unrestricted
use of the First-Order Logic (FOL).
Conclusions and future work
• The approach is applicable to semantic matching of
web services in bioinformatics, where about 3000
heterogenious services are used by scientists [3].
• We are currently implementing this matchmaking
algorithm in a public registry of biomedical services.
• In future, we plan to investigate applicability of the
approach to semantic description and retrieving of
workflows (i.e., compositions of web services).
Workflows are commonly used by scientists
conducting experiments on genomic data.
REFERENCES:
[1] D.Martin et al. Bringing Semantics to Web Services: The OWL-S Approach. In Proc. of SWSWPC’04. 2004
[2] Duncan Hull, Evgeny Zolin, Andrey Bovykin, Ian Horrocks, Ulrike Sattler, Robert Stevens. Deciding Semantic
Matching of Stateless Services. In Proc. of AAAI’06, Boston,USA, July 16-20, 2006
[3] D.Hull, K.Wolstencroft, R.Stevens, C.Goble, M.Pocock, P.Li, T.Oinn. Taverna: A tool for building and
running workflows of services. In Nucleic Acids Research, 34:W729-W732 (Web Server Issue), 2006
Acknowledgements and links:
The work is supported by EPSRC,
grants GR/S63168/01, GR/R67743/01
DynamO: http://dynamo.man.ac.uk
myGrid:
http://www.mygrid.org.uk