Management of heterogeneity in the Semantic Web

1
Management of heterogeneity in the Semantic Web
Semantic Web and Databases Atlanta, Georgia,
USA April, 2006

• Paolo Atzeni Pierluigi Del Nostro

2
Semantic Web

• Many languages and models exist
• Interoperability is the challenge, with a generic
  approach

3
Semantic Web tecnologies

• Two families of standards

ISO
W3C
OWL Constraints TMCL
RDFSchema Constraints TMCL
RDF Data models Topic maps
RDF/XML, N3 Syntaxes XTM, HyTM, LTM
4
Outline

• RDF and Topic maps what they share
• RDF definition
• Topic maps definition
• RDF vs Topic maps differences
• Model independent approach
• Meta-constructs
• Super model
• Translation process

5
RDF and Topic Maps

• Both RDF and Topic Maps
• consist of standard family
• attempt to apply knowledge representation
  techniques to information management
• define abstract models and interchange syntaxes
  based on XML
• have models that are simple and elegant at one
  level but extremely powerful at another

6
RDF (Resource Description Framework)

• Based on three concepts
• Resource all it is possible to describe. Each
  resource is identified by an URI (not only web
  resources)
• Property an attribute associated with the
  resource.
• Statement all is it possible to say about
  resources. It has the form of the triple
  ltsubject, predicate, objectgt where
• Subject resource
• Predicate property
• Object resource/literal

7
RDF (example)
8
Topic Maps

• A standard for defining knowledge structures and
  associating them with information resources
• Topic maps are organized around the concept of
  Topic, which is used to represent some real-world
  thing
• Three constructs are provided for describing the
  subjects represented by the topics
• Names multiple base names to a single topic and
  variants of each base name
• Occurrences a topic may be linked to one or more
  information resources that are deemed to be
  relevant to the topic
• Associations have a type, can be n-ary and each
  topic participate with a specific role

9
Topic Maps (example)
occurrence
occurrence
Cats are furry carnivorous animals
cats.doc
Birds are feathery animals
birds.doc
10
RDF vs Topic Maps
RDF Differences Topic maps
formal logic and mathematical graph theory roots finding aids indexes, glossaries, thesauri
machines perspectives humans
resource-centric points of view subject-centric
"lower-level" levels of semantic "higher-level"
addressable by URI subjects may be addressable or not
binary, have direction assertions n-ary, bidirectional, participants with roles
11
Our approach

• Translation between Semantic Web models handled
  with a metamodel tecnique developed for
  translating schemas from a datamodel to another

12
Constructs a classification

• Lexical types
• Sets of printable values
• The domain
• Abstract types
• Entity type, set of objects in the world
• Class, set of objects in the system
• Aggregation
• a construction based on (subsets of) cartesian
  products
• Relationship in the E-R model
• Relation in the relational model
• Function
• Attribute in the E-R model
• Function in a functional data model
• Grouping
• Hierarchies

• A model can be defined in terms of the
  meta-constructs its constructs refer to
• E.g., the E-R model
• Abstract (called Entity)
• Function from Abstract to Lexical (Attribute)
• Aggregation of abstracts (Relationship)

13
 The supermodel

• A model that includes all the meta-constructs (in
  their most general forms)
• Each model is subsumed by the supermodel
• Each schema for any model is also a scheme for
  the supermodel
• Translations are realized within the supermodel
• It needs to be extended to properly represent
  Semantic Web formalisms
• The separation between schemas and instances is
  not strong

14
The translation process
Super model
SM_RDF
SM_TM
correspondence
correspondence
translation
SM_S1
SM_S2
source
source
target
copy
copy
?
15
 The extended supermodel

• SM_Abstract(schemaOID, abstractOID, name,
  class/instance, isProperty)
• SM_Collection(schemaOID, collectionOID, name,
  type)
• SM_ComponentOfCollection(schemaOID, componentOID,
  name, objectOID, collectionOID, position,
  lexicalValue)
• SM_Identity(schemaOID, identityOID, name, type,
  value, objectOID, idObjectOID, idAssertionOID)
• SM_AttributeOfAbstract(schemaOID, attributeOID,
  name, subjectOID, predicateOID, objectOID)
• SM_AggregationOfAbstract(schemaOID,
  aggregationOID, name)
• SM_ComponentOfAggregation(schemaOID,
  componentOID, name, aggregationOID, roleOID,
  memberOID)
• SM_Assertion(schemaOID, assertionOID, name)
• SM_InstanceOf(schemaOID, attributeOID, name,
  instanceOID, classOID)
• SM_SubClassOf(schemaOID, attributeOID, name,
  subclassOID, superclassOID)
• SM_Scope(schemaOID, attributeOID, name,
  assertionOID, scopeOID)
• SM_Type(schemaOID, attributeOID, name,
  assertionOID, typeOID)
• SM_AssertionAboutAssertion(schemaOID, assOID,
  name, assSubjOID, objectOID, lexicalValue)
• SM_Domain(schemaOID, attributeOID, name,
  propertyOID, domainOID)
• SM_Range(schemaOID, attributeOID, name,
  propertyOID, rangeOID)

16
The translation process
Super model
SM_RDF
SM_TM
correspondence
correspondence
translation
SM_S1
SM_S2
source
source
target
copy
copy
?
17
  Correspondences
RDF to SM
TM to SM
18
The translation process
Super model
SM_RDF
SM_TM
correspondence
correspondence
translation
SM_S1
SM_S2
source
source
target
copy
copy
?
19
Translation rules

• Datalog variant with
• OID invention, Skolem functions are used to
  generate new identifiers when needed
• Elementary rules are composed in order to obtain
  complex translation.

SM_AttributeOfAbstract ( sOID, assOID(cOID1,
cOID2), N(agN, cN1, cN2), null, mOID1, mOID2) ?
SM_AggregationOfAbstract(sOID, agOID,
agN), SM_ComponentOfAggregation(sOID, cOID1, cN1,
agOID, mOID1, rOID1), SM_ComponentOfAggregation(sO
ID, cOID2, cN2, agOID, mOID2, rOID2), cOID1ltgtcOID2

20
Schema representation inside the Super model
type
RDF_AllResources
RDF_Resources
RDF_Statement
21
Schema representation inside the Super model
type
22
Conclusions

• Model independent approach to the translation
• Thought for database models
• Extended to embody Semantic Web formalisms
• Work in progress
• Currently developing the details of the
  translations by using the prototype ModelGen

23
Management of heterogeneity in the Semantic Web
Semantic Web and Databases Atlanta, Georgia,
USA April, 2006
Thank you

• Paolo Atzeni Pierluigi Del Nostro