Contextaware, Ontology based, Semantic Service discovery

1
Context-aware, Ontology based, Semantic Service
discovery

• Tom Broens
• 2nd of July 2004

Graduation Committee University of Twente Dr.
ir. M.J. van Sinderen P. Dockhorn Costa,
MSc. Telematica instituut Dr. ir. J.W.
Koolwaaij S.V. Pokraev, MSc.
2
Content

• Introduction to Service discovery
• Approach
• Problem analysis
• COSS approach
• Demonstration
• Conclusions
• Questions

3
Introduction to Service discovery

• Growing amount of E-services
• Services are heterogeneous and distributed
• Before a user can use a service he has to find
  it!
• Service discovery is the act of finding services
• Difficult task!

4
Scenario

• User is looking for a webshop that sells CDs
• Syntactic matching ? poor result
• Synonyms (CD Compact Disc)
• Homonyms (Shop ? Webshop ManufacturingShop)
• What do you mean? ? semantic reasoning
• Implicit information ? context
• Preferences
• Opening time ? open shops
• Location ? shops nearby

Improve service discovery mechanisms to deliver
higher quality results
5
Approach

• Problem analysis
• Approach development
• Evaluation

6
Service discovery

• Recall / Precision
• Four main retrieval technologies
• Current keyword/table
• No context
• No semantic reasoning
• No common understanding

7
CA service discovery

• CA service discovery incorporates context into
  the service discovery process
• E.g. time, location, weather forecast, shopping
  list
• Request completion ? higher precision
• Input completion ? higher recall
• Still
• No common understanding
• No semantic reasoning

8
Ontology based service discovery

• Ontologies a formal, explicit definition of a
  shared conceptualization
• Common understanding
• Mathematical approach
• Trend ? OWL W3C recommendation
• Semantic reasoning
• Issues
• Language
• Environment

9
WASP services platform

• Developed by TI, UT, Ericsson
• Three main components
• Good opportunity to apply novel approach
• Intelligent service discovery
• CA platform
• Uses ontologies
• Limited service discovery
• Limited discovery algorithm
• No CA service discovery

10
Problem analysis summary
11
COSS approach

• Starting point
• Context-aware
• Ontologies
• Semantic reasoning
• Model properties

CONTEXT
ONTOL.
SR
COSS
SD
Result
In Visa
In CreditCard
SD
SR
Out Music
Out CD
Atr Open
Atr Open time
12
Matching degree (1/2)

• Service discovery has to return a close match
  among users request and service advertisement.
• What is close ? Matching degree ? quality of
  the match

Shop
Music
Nearby
Open
13
Matching degree (2/2)

• Types of matches
• COSS matches
• Precise (1, 2)
• Approximate (3, 4)
• Mismatch (5)

14
COSS algorithm

• Four step algorithm
• Service type ? hard criteria (Shop)
• Outputs ? hard criteria (Music)
• Inputs ? soft criteria (Cash)
• Attributes ? soft criteria (Nearby, Open)

15
COSS attribute model

• Concept lattices
• Formal concept analysis
• Find natural clusters of objects and their
  properties in an input data set
• Triple lattice(o,a,r)

Generic
Specific
16
Lattices in COSS

• Service request is one object in the table
• Services are objects in the table
• Service attributes are columns
• Relations indicated by crosses in the table

Lattice table
Lattice diagram
17
Demonstration
18
Evaluation

• Reflection on issues from the problem analysis
• Synonyms
• Homonyms
• Context
• Common understanding
• Semantic reasoning
• Query evaluation
• Every query showed an improved recall/precision
  rate
• Query, Advertisements and ontology adapted to
  each other

19
Conclusions

• Novel service discovery approach COSS
• Context-awareness
• Ontologies
• Semantic reasoning
• Prototype
• Integration in WASP services platform with
  webservices
• Higher quality discovery results
• Basis for future developments
• COSS applied in a delimited domain

20
Future work

• Performance Scalability
• Usage of context
• Ontology maintenance
• Ontology relations
• Human Computer Interaction
• Personalization

21
?
Questions
22
(No Transcript)