METEOR-S%20WEB%20SERVICE%20ANNOTATION%20FRAMEWORK%20(MWSAF)

1
METEOR-S WEB SERVICE ANNOTATION FRAMEWORK(MWSAF)
Abhijit Patil, Swapna Oundhakar, Amit Sheth,
Kunal Verma LSDIS Lab, Department of Computer
Science,The University of Georgia
2
Outline

• Introduction
• METEOR-S Project _at_ UGA
• SchemaGraph
• Architecture
• Matching algorithm
• Results
• Conclusions and Future Work

3
Introduction

• Semantic Web Services
• Explicate semantics of the Web service provider
• Use existing domain ontologies to provide
  contextual normalization
• Challenges
• Finding relevant domain ontologies
• Finding appropriate concepts in the ontologies
• Need a tool for allowing semi-automatic
  annotation

4
Semantic Web services

• Describe services with ontology based languages
  e.g. OWL-S
• Add semantics to existing Web service standards
  e.g. METEOR-S
• Common factor

OWL-S Describe Web services using ontology based
service description languages
METEOR-S Add Semantics by adding annotations to
service descriptions in WSDL
Common Factor Relate Web service I/O parameters
with Ontological concepts
5
METEOR-S Web service Annotation

• Map Web services inputs and outputs data
  represented using XML schema to concepts in
  ontologies
• Annotate WSDL with Ontologies
• How ?
• Borrow from Schema matching
• Semantic disambiguation between terms in XML
  messages represented in WSDL and concepts in
  ontology
• Match XML schema elements from WSDL to
  ontological concepts

6
METEOR-S Web Service Annotation Framework (MWSAF)

• Assumptions
• Domain is depicted by one or more domain
  ontologies
• A Web service may belong to one or more domains
• MWSAF Functionality
• Find the domain(s) of the Web service
• Annotate the Web service with one or more
  ontologies

7
Outline

• Introduction
• METEOR-S Project _at_ UGA
• SchemaGraph
• Architecture
• Matching algorithm
• Results
• Conclusions and Future Work

8
METEOR-S Project _at_ UGA

• METEOR-S exploits Workflow, Semantic Web, Web
  Services, and Simulation technologies to meet
  these challenges in a practical and standards
  based approach.
• Applying Semantics in Annotation, Quality of
  Service, Discovery, Composition, Execution of Web
  Services
• Adding semantics to different layers of Web
  services conceptual stack
• Use of ontologies to provide underpinning for
  information sharing and semantic interoperability

http//swp.semanticweb.org, http//lsdis.cs.uga.ed
u/proj/meteor/swp.htm
9
Semantics in METEOR-S and WS stack
MWSCF Semantic Web Process Composition Framework
Flow
MWSDI Scalable Infrastructure of Registries for
Semantic publication and discovery of Web Services
Description
MWSAF Semantic Annotation of WSDL (WSDL-S)
Messaging
Network
METEOR-S at the LSDIS Lab exploits Workflow,
Semantic Web, Web Services, and Simulation
technologies to meet these challenges in a
practical and standards based approach
http//swp.semanticweb.org, http//lsdis.cs.uga.ed
u/proj/meteor/swp.htm
10
METEOR-S Types of Semantics

• Data / Information Semantics
• What Formal definition of data in input and
  output messages of a web service
• Why For Discovery and Interoperability
• How By annotating input/output data of web
  services using ontologies
• Functional Semantics
• What Formally representing capabilities of web
  service
• Why For Discovery and Composition of Web
  Services
• How By annotating operations of Web Services as
  well as provide preconditions and postconditions

11
METEOR-S 4 types of Semantics

• Execution Semantics
• What Formally representing the execution or
  flow of services in a process or operations in a
  service
• Why For Analysis (verification), Validation
  (simulation) and Execution (exception handling)
  of the process models
• How Using State Machines, Petri nets, activity
  diagrams etc.
• QoS Semantics
• What Formally describing operational metrics of
  a web service/process
• Why To select the most suitable service to
  carry out an activity in a process
• How Using QoS model Cardoso and Sheth, 2002
  for web services

12
METEOR-S Architecture
13
WSDL-S Metamodel
14
WSDL-S

• lt?xml version"1.0" encoding"UTF-8"?gt
• ltdefinitions
• name "BatterySupplier"
• targetNamespace "http//lsdis.cs.uga.edu/meteor
  /BatterySupplier.wsdl20"
• xmlns "http//www.w3.org/2004/03/wsdl"
• xmlnstns "http//lsdis.cs.uga.edu/BatterySuppl
  ier.wsdl20"
• xmlnsrosetta " http//lsdis.cs.uga.edu/project
  s/meteor-s/wsdl-s/pips.owl "
• xmlnsmephttp//www.w3.
  rosettaPurchaseOrderStatusResponse
  org/TR/wsdl20-patternsgt
• ltinterface name "BatterySupplierInterface"
  description "Computer PowerSupply Battery Buy
  Quote Order Status "
• domain"naicsComputer and
  Electronic Product Manufacturing" gt
• ltoperation name "getQuote" pattern
  "mepin-out" action "rosettaRequestQuote" gt
• ltinput messageLabel qRequest element
  "rosettaQuoteRequest" /gt
• ltoutput messageLabel quote
  element "rosettaQuoteConfirmation" /gt
• lt/operationgt
• ltoperation name "placeOrder" pattern
  "mepin-out" action "rosettaRequestPurchase
  Order" gt
• ltinput messageLabel order element
  "rosettaPurchaseOrderRequest" /gt
• ltoutput messageLabel
  orderConfirmation element "rosettaPurchaseO
  rderConfirmation" /gt

P. Rajasekaran, J. Miller, K. Verma, A. Sheth,
Enhancing Web Services Description and Discovery
to Facilitate Composition, available at
http//lsdis.cs.uga.edu/lib/download/swswpc04.doc
15
Outline

• Introduction
• METEOR-S Project _at_ UGA
• SchemaGraph
• Architecture
• Matching algorithm
• Results
• Conclusions and Future Work

16
Matching Issues (WSDL and Ontologies)

• Expressiveness
• Different reasons behind their development
• XML Schema used in WSDL for providing basic
  structure to data exchanged by Web services
• Ontologies are developed to capture real world
  knowledge and domain theory
• Knowledge captured
• XML Schema has minimal containment relationship
• Language used to describe ontologies model real
  world entities as classes, their properties and
  provides named relationships between them
• Solution
• Use hueristics to create normalized
  representation
• We call it SchemaGraph

17
MWSAF SchemaGraph

• What is SchemaGraph ?
• Normalized representation to capture XML Schema
  and DAML Ontology
• How to use SchemaGraph
• Conversion functions convert both XML Schema and
  Ontology to SchemaGraph representation
• XML schema used by WSDL ? W wc1, wc2, wc3, ,
  wcn where, wci is an element in XML schema and n
  is the number of elements
• Ontology ? O oc1, oc2, oc3, , ocm where, oci
  is a concept in Ontology and m is the number of
  concepts
• Match function takes both W and O and returns a
  set of mappings

18
MWSAF XML Schema to SchemaGraph
Rule XML Schema constructs SchemaGraph representation
1 Element, Node
2 simpleType Node
3 Enumeration values defined for simpleType S Node with edge between simpleType S node and value node with name hasValue
4 ComplexType Node
5 Sub-elements of complexType C which have range as basic XML datatypes Node with edge between complexType C node and this node with name hasElement
6 Sub-elements of complexType C which have range as complexTypes or simpleTypes or elements defined in same schema Edge between complexType C node and the range type node
19
MWSAF XML Schema to SchemaGraph
- ltxsdcomplexType name"WeatherReport"gt -
ltxsdsequencegt   ltxsdelement
name"phenomena" type"xsd1Phenomenon" /gt  
ltxsdelement name"wind" type"xsd1Wind" /gt  
lt/xsdsequencegt   lt/xsdcomplexTypegt -
ltxsdcomplexType name"Phenomenon"gt -
ltxsdsequencegt   ltxsdelement name"type"
type"xsd1PhenomenonType" /gt   ltxsdelement
nameintensity" type"xsd1PhenomenonIntensity"
/gt   lt/xsdsequencegt   lt/xsdcomplexTypegt -
ltxsdcomplexType name"Wind"gt -
ltxsdsequencegt   ltxsdelement
name"gust_speed" type"xsddouble" /gt  
ltxsdelement name"prevailing_direction"
type"xsd1Direction" /gt   lt/xsdsequencegt  
lt/xsdcomplexTypegt - ltxsdsimpleType
name"PhenomenonType"gt - ltxsdrestriction
base"xsdstring"gt   ltxsdenumeration
value"MIST" /gt   ltxsdenumeration
value"FOG" /gt   ltxsdenumeration
valueSNOW" /gt   ltxsdenumeration
value"DUST" /gt   lt/xsdrestrictiongt  
lt/xsdsimpleTypegt
WeatherReport
wind
phenomena
Wind
Phenomenon
hasElement
prevailing_direction
type
Direction
gust_speed
intensity
PhenomenonType
Rule 1 Element gt Node
PhenomenonIntensity
hasValue
hasValue
hasValue
hasValue
MIST
FOG
SNOW
DUST
20
MWSAF - Ontology to SchemaGraph
Rule Ontology representation SchemaGraph representation
1 Class Node
2 Property of class D with basic datatype as range (Attribute) Node with edge joining it to class D node with name hasProperty
3 Property of class D with other class R as range (Relation) Edge between class D node and range class R node
4 Instance of class C Node with edge joining class C node to instance node with name hasInstance
5 Class(X)-subclass(Y) relationship Edge between class X node and class Y node with name hasSubclass
21
MWSAF - Ontology to SchemaGraph
- ltdamlClass rdfID"WeatherPhenomenon"gt  
ltrdfscommentgtSuperclass for all weather
eventslt/rdfscommentgt   ltrdfslabelgtWeather
eventlt/rdfslabelgt   lt/damlClassgt - ltdamlClass
rdfID"WindEvent"gt   ltrdfssubClassOf
rdfresource"WeatherPhenomenon" /gt  
lt/damlClassgt - ltdamlProperty rdfID"windDirecti
on"gt   ltrdfsdomain rdfresource"WindEvent"
/gt   lt/damlPropertygt - ltdamlClass
rdfID"GustingWindEvent"gt   ltrdfssubClassOf
rdfresource"WindEvent" /gt   lt/damlClassgt -
ltdamlClass rdfID"CurrentWeatherPhenomenon"gt  
ltrdfssubClassOf rdfresource"WeatherPhenomenon
" /gt   lt/damlClassgt - ltdamlClass
rdfID"OtherWeatherPhenomena"gt  
ltrdfssubClassOf rdfresource"CurrentWeatherPhen
omenon" /gt   lt/damlClassgt - ltdamlClass
rdfID"Duststorm"gt   ltrdfssubClassOf
rdfresource"OtherWeatherPhenomena" /gt  
lt/damlClassgt - ltdamlClass rdfID"PrecipitationE
vent"gt   ltrdfssubClassOf rdfresource"Current
WeatherPhenomenon" /gt   lt/damlClassgt -
ltdamlClass rdfID"SolidPrecipitationEvent"gt  
ltrdfssubClassOf rdfresource"PrecipitationEvent
" /gt   lt/damlClassgt - ltdamlClass
rdfID"Snow"gt   ltrdfssubClassOf
rdfresource"SolidPrecipitationEvent" /gt  
lt/damlClassgt - ltdamlClass rdfID"ObscurationEve
nt"gt   ltrdfssubClassOf rdfresource"CurrentWe
atherPhenomenon" /gt   lt/damlClassgt -
ltdamlClass rdfID"Fog"gt   ltrdfssubClassOf
rdfresource"ObscurationEvent" /gt  
lt/damlClassgt - ltdamlClass rdfID"Mist"gt  
ltrdfssubClassOf rdfresource"ObscurationEvent"
/gt   lt/damlClassgt
WeatherPhenomenon
WindEvent
Rule 1
Rule 5
windDirection
GustingWindEvent
Rule 2
CurrentWeatherPhenomenon
OtherWeatherPhenomenon
Duststorm
PrecipitationEvent
ObsucurationEvent
SolidPrecipitationEvent
Snow
Mist
Fog
22
Outline

• Introduction
• METEOR-S Project _at_ UGA
• SchemaGraph
• Architecture
• Matching algorithm
• Results
• Conclusions and Future Work

23
MWSAF Architecture

• Ontology Store
• Categorize in domains
• Currently supports DAML and RDF formats
• Will be replaced in future with high quality
  ontology search mechanisms
• Parser Library
• Parser used to generate SchemaGraphs
• Currently provides Ontology2Graph and WSDL2Graph
  parsers
• Matcher Library
• Provides two types of Matching algorithms
• Element level Matching algorithms NGram,
  CheckSynonyms, CheckAbbreviations, TokenMatcher
• Schema Matching algorithms
• Allows to add new Algorithms
• User Interface
• Displays the mappings and allows user to accept
  or reject it
• It also allows to match the concepts manually
• Displays the WSDL and ontology in tree format

24
MWSAF Architecture
SchemaGraph For WSDL
SchemaGraph For Ontology
getBestMapping (Ranking algorithm)
WSDL Concept Ontology Concept Match Score
Phenomenon WeatherEvent 0.51
windEvent Wind 0.79
Annotated WSDL file
25
Outline

• Introduction
• METEOR-S Project _at_ UGA
• SchemaGraph
• Architecture
• Matching algorithm
• Results
• Conclusions and Future Work

26
MWSAF Matching two concepts

• IOParametersMatch (w,o)
• ElemMatch (w,o) SchemaMatch (w,o)
• ElemMatch (w,o) gt Element level match
• SchemaMatch (w,o) gt Schema level match
• subTree(w) subTree(o)

FUNCTION findMapping
INPUT wci ? W , oci ? O
OUTPUT mi ( wci, oci, MS )
27
MWSAF Element level Match

• Definition
• Element level match is the measure of the
  linguistic similarity between two concepts based
  on their names.
• Assumption Concepts from XML schema and
  ontology have meaningful names
• ElemMatch (w,o) gt Element level match
• NameMatch with stemming
• SynonymsMatch Snow and snowFall mean the same
• HypernymRelation (w is a kind of o)
  prevailing_speed is a type of speed of a wind
  i.e. windSpeed
• HyponymRelation (o is a kind of w)
• Acronyms Sea Level Pressure has acronym SLP

28
MWSAF Element level Match (example)
29
MWSAF Element level Match

• Element level match algorithms used by MWSAF
• NGram This algorithms calculates similarity
  between two strings by considering the number of
  qgrams that they have in common. It uses dice
  coefficient to calculate this similarity.
• CheckSynonyms This algorithm uses WordNet to
  find synonyms. It also accounts for hypernyms and
  hyponyms matching.
• CheckAbbreviation This algorithm uses domain
  specific Abbreviation dictionary to expand the
  abbreviations
• TokenMatcher This algorithm uses the Porter
  Stemmer to find the roots of the words. It also
  uses tokenization based on punctuation and
  capitalization of letters.

30
MWSAF Element level Match
where, ms1 MatchScore ( NGram ) ms2
MatchScore ( Synonym Matching ) ms3 MatchScore
( Abbreviation Expansion ) ms4 MatchScore (
Token Matching )
Example
WSDL Concept Ontological Concept ElemMatch Algorithm
wind WindEvent 0.639 NGram
wind WindChill 0.478 NGram
snow SnowFall 1 Synonyms
slp SeaLevelPressure 1 Abbreviation
relative_humidity RelativeHumidity 1 NGram
31
MWSAF Schema level Match

• Definition
• The Schema level match is the measure of
  structural similarity between two concepts
• It is based on sub-concept similarity
  (subConceptSim) and sub-concept match
  (subConceptMatch).

32
MWSAF Schema level Match

• Definition Sub-concept Similarity (
  subConceptSim )
• The sub-concept similarity is the average match
  score of each individual sub-element of the
  concept

• Definition Sub-concept Match ( subConceptMatch
  )
• The sub-concept match is the fraction of total
  number of sub-elements of a concept that are
  matched

33
MWSAF Schema level Match (example)
Example
WSDL Concept Pressure Ontological Concept PressureEvent MS
delta ---- 0
slp Sea Level Pressure 1
relative_humidity RelativeHumidity 1
subConceptSim ( Pressure , PressureEvent ) ( 1 1 0 ) / 3 0.667 subConceptSim ( Pressure , PressureEvent ) ( 1 1 0 ) / 3 0.667 subConceptSim ( Pressure , PressureEvent ) ( 1 1 0 ) / 3 0.667
subConceptMatch ( Pressure , PressureEvent ) 2 / 3 0.667 subConceptMatch ( Pressure , PressureEvent ) 2 / 3 0.667 subConceptMatch ( Pressure , PressureEvent ) 2 / 3 0.667
34
MWSAF Categorizing WSDL

• Average Service Match ( avgServiceMatch )
• Calculated as the average match of all the
  concepts of a WSDL schema and a domain ontology
• The domain of the ontology corresponding to the
  best average service match also represents the
  domain of the Web service
• Normalized on the scale of 0 to 1

where, k number of mapped concepts n number
of concepts in WSDL schema
35
MWSAF Annotating WSDL

• Average Concept Match ( avgConceptMatch )
• Calculated as the average match of the mapped
  concepts of a WSDL schema
• Based on this measure user can decide whether to
  accept mappings for annotation or not
• It is normalized on the scale of 0 to 1

36
Outline

• Introduction
• METEOR-S Project _at_ UGA
• SchemaGraph
• Architecture
• Matching algorithm
• Results
• Conclusions and Future Work

37
MWSAF Categorizing WSDL

• 6 different Web services are compared to 5
  ontologies to get avgServiceMatch values for each
  of them. Service belongs to the domain of the
  ontology for which it gives best avgServiceMatch.
• E.g. AirportWeather service best matches to
  weather-ont ontology and hence belongs to weather
  domain

38
MWSAF Categorizing WSDL

• 24 Web services from Weather and Geographical
  domain are categorized with different threshold
  (CT) values.
• For CT 0.4, two services are categorized
  wrongly
• For CT 0.5, all the Web services are not
  categorized

39
MWSAF Testing

• With original Geo ontologies, services gave low
  match scores
• By adding few more concepts, the match scores
  improved for many services.
• Plot of number of mapped concepts strengthens
  this observation

40
MWSAF Testing

• Problems
• Match scores are low
• All concepts are not mapped
• Reasons
• Match algorithms can be improved
• Domain specific synonyms and abbreviations can
  improve avgConceptMatch
• Domain specific match algorithms can be
  implemented
• Ontologies are still in development stage and not
  comprehensive enough to contain all the concepts
  from the domain
• Need ontologies specifically designed for Web
  services
• WSDL files are automatically generated by web
  servers and hence not all IO parameters have
  meaningful names

41
Outline

• Introduction
• METEOR-S Project _at_ UGA
• SchemaGraph
• Architecture
• Matching algorithm
• Results
• Conclusions and Future Work

42
Conclusions and Future Work

• Conclusions
• Created an initial prototype for semi-automatic
  annotation of Web services
• Initial results promising, but a lot of
  improvement possible
• WSDL-S adds semantics to Web services with
  minimal changes
• Future Work
• Apply machine learning techniques to improve
  accuracy
• Build a test bed for Semantic Web Services
• Eclipse based tool release in 1 montg

43
MWSAF Screenshot
44
References

1. D. Fensel, C. Bussler, "The Web Service Modeling
   Framework WSMF", Technical Report, Vrije
   Universiteit Amsterdam
2. METEOR-S Semantic Web Services and Processes,
   http//swp.semanticweb.org
3. A. Ankolekar, M. Burstein, J. Hobbs, O. Lassila,
   D. Martin, D. McDermott, S. McIlraith, S.
   Narayanan, M. Paolucci, T. Payne, and K. Sycara,
   "DAML-S Web service Description for the Semantic
   Web," in Proceedings of the 1st International
   Semantic Web Conference (ISWC 2002)
4. S. Agarwal, S. Handschuh, and S. Staab Surfing
   the Service Web, in Proceedings of the 2nd
   International Semantic Web Conference (ISWC 2003)
5. M. Klein, Combining and relating ontologies an
   analysis of problems and solutions. in (IJCAI
   2001)
6. E. Rahm and P. A. Bernstein. A survey of
   approaches to automatic schema matching. In The
   VLDB Journal Volume 10 Issue , (2001), pages
   334-350, 2001.
7. H. Do, S. Melnik, and E. Rahm. Comparison of
   schema matching evaluations. In Proceedings of
   the 2nd Int. Workshop on Web Databases (German
   Informatics Society), 2002
8. Pottinger, R. A. and P. A. Bernstein, Merging
   Models Based on Given Correspondences. Proc.
   29th VLDB Conference