Languages for the Semantic Web and Semantic Web Services

1
Languages for the Semantic Web and Semantic Web
Services

• Current Efforts and Challenges for ASP
• Axel Polleres
• axel.polleres_at_deri.org

2
Semantic Web

• Publish machine-processable meta-data on the Web
  (semantic Web idea!)!
• Provide the means to publish data on relations of
  resources and taxonomies of data on the Web
• Provide standards on top of XML to describe the
  meaning of published knowledge
• This meta-data shall ideally be consensual
  (Ontologies!).

3
Overview

• OWL ? Semantic Web
• RDF(S), OWL, SWRL, SWRL FOL
• Semantic Web Services
• SWSL-Rules
• WSML
• Challenges

4
OWL ? Semantic WebRDF(S), OWL, SWRL, SWRL FOL

• RDF(S) W3C Recommendation, latest version,10
  February 2004
• simple taxonomies, express structured knowledge
  in a graph made up of ltsubject predicate objectgt
  triples
• OWL W3C Recommendation, latest version,10
  February 2004
• SWRL W3C member submission, 21 May 2004
• extends OWL DL by simple rules, quite
  restrictive, but undecidable already.
• This is not all!
• Several drawbacks, several alternative
  suggestions. Let's see the results of the current
  Rules Workshop going on this minute in Washington.

5
RDFS - OWL Lite/DL/Fullinappropriate Layering

• OWL DL is not properly layered on top of RDFS
  syntactically
• even worse Given the same (OWL DL) ontology
  there are things semantically entailed by OWL
  Full, not entailed by OWL Lite
• something's strange here!

6
OWL DL is not properly layered on top of RDFS
syntactically

• OWL does not smoothly integrate
• No meta-statements in OWL Lite and DL (i.e.
  separate vocabulary
• RDF Triples
• hansi rdftype eagle.
• eagle rdftype species.
• Not possible in OWL Lite and DL!

7
Given the same (OWL DL) ontology there are things
semantically entailed by OWL Full, not entailed
by OWL DL

• From OWL Semantics and abstract Syntax document
• The "only if" direction is not true!
• I.e., OWL Full adds additional inferences on the
  same OWL DL ontology!

Theorem 2 Let O and O' be collections of OWL DL
ontologies and axioms and facts in abstract
syntax form that are imports closed, such that
their union has a separated vocabulary. Given a
datatype map D that maps xsdstring and
xsdinteger to the appropriate XML Schema
datatypes and that includes the RDF mapping for
rdfXMLLiteral, then the translation of O OWL
Full entails the translation of O' with respect
to D if the translation of O OWL DL entails the
translation of O' with respect to D.
8
OWL DL entails less than OWL full
Axel friendOf Lea . ² OWL Full Axel rdftype x
. x owlonProperty friend . x
owlminCardinality "1"xsdnonNegativeInteger
. But Axel friendOf Lea . ² OWL DL Axel
rdftype x . x owlonProperty friend . x
owlminCardinality "1"xsdnonNegativeInteger .
Woulld need additionally Axel rdftype
owlThing. Lea rdftype owlThing . friend
rdftype owlObjectProperty .
9
SWRL

• OWL unsatisfactory expressivity, not even simply
  rules like
• parent(?x,?y) ? brother(?y,?z) ? uncle(?x,?z)
• SWRL extends OWL with a simple rule language
  which allows concept (unary) and role (binary)
  predicates as well as sameAs (equality) and
  differentFrom (inequality) as atoms.
• Undecidable, if not further restricted (e.g. to
  Description logic programs)
• SWRL FOL W3C member submission April 11th
• Proposal to loosen some of the restrictions of
  SWRL to handle function-free handle unary/binary
  first-order logic, n-ary predicates only via
  reification.

10
RDF(S), OWL, SWRL syntax

• All of these standards offer XML exchange
  syntaxes
• RDF exchange syntaxes
• OWL/RDF, SWRL/RDF exchange syntax are not really
  useful puts OWL constructs inside an RDF graph
  themselves, causes problems with OWL DL and OWL
  Full semantic interoperability

11
Semantic Web Services- Requirements

• Semantic descriptions of the functional and
  behavioral aspects of Web Services to enable
  (semi-)automatic discovery, composition and
  execution of Web Services, build upon Web
  Services technologies (SOAP, WSDL, UDDI).
• pre-postconditions
• Interfaces
• Mediators
• Capability
• User goals, preferences
• Currently three major initiatives
• OWL-S an OWL ontology to semantically describe
  Web Services (OWL)
• WSMO/WSML/WSMX a general framework for SWS
  description (own logical language)
• SWSL a framework for SWS description based on
  OWL (but also own language)
• Another W3C workshop in June in Innsbruck
• Frameworks for Semantic Web Services deadline
  for position papers extended to Friday April 29th!

12
Why OWL is not enough for SWS?

• Permit FOL for ontologies beyond DL/OWL Need to
  express complex conditions, rules, trust policies
  for contracting, nonmon. features,
  prioritization, dynamics
• Integrate nonmon., frame/OO, DDB ontologies with
  mon. DL/FOL ontologies
• Cope robustly with conflict between ontologies,
  e.g., merging OWL ontologies from many sources
• OWL-S does not define the semantics of dynamics,
  proposes to allow DRS,KIF,SWRL for expressing
  pre-conditions, effects, etc. but does not
  specify the semantics.

13
SWSL-Rules

• Semantic Web Service Language initiative
• (Micheal Kifer, David Martin, Benjamin Grosof,
  DAML founded plus European participants)
• Ontology/Rules Language
• SWSL Rules LP with NAF Courteous, Hilog
  extensions
• SWSL FOL
• Shared presentation syntax builds on F-Logic
• Markup syntax TBD probably with RuleML
  committee
• W3C member submission planned

14
SWSL Language Layers

• Strong Consensus Need Nonmonotonic LP. And
  FOL.
• SWSL-Rules the LP KR.
• SWSL-FOL the FOL KR.
• Expressive Features for SWSL are similar to those
  desired for SW rules in general, but with bit
  different near-term importance/urgency
• Important in both Prioritization, NAF (cf.
  Courteous LP)
• Important in both, more urgent in SWS than SW
  overall Meta- power/convenience Hilog, frame
  syntax (cf. F-Logic)
• A bit more important in SWS than SW overall
  Lloyd-Topor (nested expressions)
• Reification meta-knowledge/modeling, mentioned
  already in RDF, but no semantics for it.

15
SWSL Rules SWSL FOL

• Semantics for rules part and FOL part separate
• Exchange syntax RuleML
• New fundamental KR theory is needed to unify
  nonmon. LP with FOL
• "A holy grail for SWS, and for SW generally"

16
WSML (joint work with Jos de Bruijn, Holger
Lausen, Dieter Fensel, Michael Kifer)

• Developed as joint effort in several EU Projects
  (dip, SEKT, KnowledgeWeb)
• Based on Web Service Modeling Ontology WSMO
• Also has its own rules language, similarities
  with SWSL
• W3C member Submission pending

17
Web Service Modeling Language

• Four elements of WSMO
• Ontologies
• Goals
• Web Services
• Mediators
• WSML provides a formal grounding for the
  conceptual elements of WSMO, based on
• Description Logics
• Deductive Databases
• First-Order Logic

18
Syntaxes for WSML

• Human-readable syntax
• Layered syntax
• Inspired by OIL/OWL and F-Logic
• Two flavors
• Conceptual syntax
• Logical Expression Syntax
• Semantics is fixed in WSML variants
• XML syntax
• RDF syntax
• Mapping to OWL

19
Variants of WSML logical languages
20
Variants of WSML (contd.)

• WSML-Core
• Based on the intersection of Description Logics
  and Datalog
• Has (frame-based) conceptual syntax and logical
  expression syntax
• WSML-Flight
• Layered on top of WSML-Core
• Basic meta-class facility
• Constraints
• Non-monotonic features (default negation)
• Provides more intuitive modeling constructs (for
  people with DB/SE background)
• Preferred ontology modeling language
• Based on Datalog with stratified negation and
  inequality

21
Variants of WSML (contd.)

• WSML-Rule
• Based on Logic Programming with default negation
  and F-Logic/HiLog syntactical extensions
• Preferred goal/web service modeling language
• WSML-DL
• Based on SHIQ
• WSML-Full
• Combining FOL with minimal models and
  non-monotonicity

22
WSML Logical Expressions

• Elements
• Function symbols (e.g. f())
• Variables (e.g. ?x)
• The syntax is based on F-Logic style molecules,
  e.g.
• Human subClassOf Animal.
• Axel memberOf Human,
• Axelname hasValue Axel Polleres.
• Predicates (e.g. distance(?x, ?y,?z))
• Logical connectives (or, and, not, impliedBy,
  equivalent, implies, forall, exists, -, !-),
• ASCII, but readable, more or less directly
  translatable to XML tags.
• Example
• ?x memberOf Human equivalent ?x memberOf Animal
  and ?x memberOf LegalAgent.

23
WSML-Core

• Allows conceptual modeling of ontologies
• Based on Description Logic Programs subset of OWL
• i.e., efficient query answering
• Should be easily adopted in existing
  implementations (e.g. XSB, OntoBroker,
  SWI-Prolog, KAON, DLV)
• Import/export OWL ontologies
• Datatype support based on XML Schema datatypes
  and OWL-E
• Expressive enough for most current ontologies
• Properly layered on top of RDF/RDFs worked upon,
  currently restrictions on the use of RDF(S)
  vocabulary.
• Can be used for limited goal/web service modeling

24
WSML-Flight

• Is an extension of WSML-Core
• Adds limited support for nominals
• Stays in LP world
• Meta-modeling
• Adds inequality (plus UNA!)
• Adds constraints (wrt. local knowledge base)
• Adds stratified non-monotonic negation
• Allows arbitrary safe Datalog rules (i.e. no
  function symbols rules must be safe)
• Language is based on Datalog with inequality,
  constraints and stratified negation

25
WSML-Rule

• Extension of WSML-Flight
• Allows unrestricted use of function symbols
• Non-stratified negation (current suggestion use
  wfs)
• Possibly other features

26
WSML-DL

• Equivalent to SHIQ with datatype extension
• Open questions
• DL epistemology?
• DL concrete syntax for logic expressions?

27
WSML-Full

• Based on a combination of First-Order Logic and
  minimal model semantics and default negation
• Unifies rule language with first-order based
  language
• Possible formalisms
• Autoepistemic Logic
• Default Logic
• Circumscription

28
WSML Goals and Web Services

• Goal / Web Service assumptions/effects and
  pre/post-conditions are defined through WSML
  logical expressions
• Logical expressions rely on ontologies
• Use of ontologies through
• Ontology import
• Mediation
• Open issue semantics of dynamics, interfaces,
  grounding

29
WSML Conclusions

• WSML is concrete language for modeling
• Ontologies
• Web Services
• Goals
• Mediators
• Variants
• WSML-Core
• WSML-Flight
• WSML-Rule
• WSML-DL
• WSML-Full
• Modular, Frame-based
• Conceptual syntax vs. Logical Expressions
• Syntaxes
• Human readable
• XML
• RDF
• Mapping to OWL

30
Summary

• The current W3C recommendations have some
  unresolved issues
• The discussion in the semantic Web community are
  often more about syntax than about semantics
• More expressivity than OWL is necessary, the
  standards discussion is not over!
• Challenges
• For the "ontological reasoning" part
• Fix semantic layering mess on top of RDF/RDFS
• Semantic Interoperability of Knowledge bases
  building on different paradigms.
• Engines Support/frontends for XSD/XPath
  built-ins, Frame-based syntax. Efficient, modular
  reasoning support.
• Implies Accept incomplete reasoning in some
  cases.
• How to unify the DL, FOL and LP worlds? What is
  the semantics of a unifying umbrella language!?!
• Reasoning with networks of ontologies, rewriting,
  etc.
• For the Web Services Part
• Formalizing dynamic aspects of Semantic Web
  Services (pre-/postc., interfaces) in a way
  interoperable with ontologies
• Idea use of action theories, action languages,
  mappings from/to process languages, etc.

Thank you! ltquestions?/gt
31
Additional slide

• Desirable features of a unified logical
  framework
• arbitrary use of neg/naf?
• Combination of wfs/sm knowledge bases?
• Nested expressions quantifiers
• Define easily checkable layered syntactical
  restrictions with increasing expressive power and
  computational properties (also WITHIN LP, tight,
  etc.) in order to support modular
  reasoner-support.