A Software Framework for Matchmaking based on Semantic Web Technology

warbarnacleΑσφάλεια

5 Νοε 2013 (πριν από 3 χρόνια και 10 μήνες)

81 εμφανίσεις

A Software Framework for Matchmaking

based on

Semantic Web Technology

Eyal Oren

DERI

2004/04/14

on the paper by Li and Horrocks

http://www.cs.man.ac.uk/~lil/papers/p815
-
li.pdf

Eyal Oren

2

Overview


Service advertisement and discovery


DAML
-
S based service ontology


working prototype


agent based (JADE)


DL Reasoner (Racer)


quite comparable with D5.1

Eyal Oren

3

Current standards


WSDL


XML grammar for describing network services


structured way of describing communication


no support for semantic description


UDDI


registry describing businesses and offered services


no description of service capabilities, only string matching


you could use DAML
-
S (or WSMO) tModels


DAML
-
S


describing properties and capabilities of web services


“can adequately represent functionalities of web services”

Eyal Oren

4


extension/application of DAML
-
S


describe specific profiles


service advertisement


service request


perform a matching of requests and
advertisements


question is whether DAML
-
S (or DAML+OIL)
is suitable/sufficient for describing web services



in wsmo we don’t (yet) limit the language expressiveness


therefore, everything is possible and nothing is (known to be) computable


we will define a decidable subset: WSML
-
DL, WSML
-
HL, WSML
-
DLP (?)

Eyal Oren

5

Example


buying PC’s

Eyal Oren

6

Example


advertisement and request


note, that for efficiency reasons, these (individual) offers and requests are expressed as special concepts

(Tbox reasoning is more efficient than ABox reasoning)

profile does not include
providedBy and requestedBy

(as DAML
-
S prescribes)

Eyal Oren

7

Matching levels

Advertisment A and Request R


Exact: A ≡ R


PlugIn:




(offering PC, looking for second
-
hand PC)


Subsume:



(offering second
-
hand, looking for PC)


Intersection:


Disjoint

Eyal Oren

8

Prototype
-

agents


Advertiser: publishes, withdraws, modifies
advertisements


Seeker: publishes requests (volatile/persistent)


Host:


manages repository of advertisements


manages persistent queries


uses Racer to compute request
-
advertisement subsumption


could compute advertisements hierarchy off
-
line


requests subsumptions easy


inserting new advertisements easy


removing advertisements harder (re
-
classifying TBox)


Eyal Oren

9

Useful for WSMO


this works, so we need to explain:

why is DAML
-
S not sufficient (not expressive) ?


look at levels of matching

(exact, plugin, subsume, etc.)


agents infrastructure (FIPA), for WSMX (??)