Presentation is loading. Please wait.

Presentation is loading. Please wait.

Inference-based Semantic Mediation and Enrichment for the Semantic Web AAAI SSS-09: Social Semantic Web: Where Web 2.0 Meets Web 3.0 March 25, 2009 Dan.

Similar presentations


Presentation on theme: "Inference-based Semantic Mediation and Enrichment for the Semantic Web AAAI SSS-09: Social Semantic Web: Where Web 2.0 Meets Web 3.0 March 25, 2009 Dan."— Presentation transcript:

1 Inference-based Semantic Mediation and Enrichment for the Semantic Web AAAI SSS-09: Social Semantic Web: Where Web 2.0 Meets Web 3.0 March 25, 2009 Dan Hunter (presenter) Basil Krikeles Fotis Barlos

2 2 Objective: Address Interoperability within and across Communities of Interest (COIs) ● Challenges – Semantic mediation is an N 2 problem for N different schemas – Handcrafting mediators is a brittle process; Schemas/ontologies evolve – Minimize information loss ● Assumptions – COIs evolve their domain ontologies – COI self-interest suggests that intra-COI mediation is handled by the community – COI self-interest should lead to relatively stable (but not fixed) published ontologies ● Approach – Combine manual and automatic generation of maps cross using semantic web technologies, thus sidestepping both the N 2 and the brittleness problems – Evolve a repository of pairwise maps for mediator generation – Infer mappings to minimize information loss RDB XML OWL other COI 1 Intra-COI mediation COI 2 Inter-COI or cross-COI mediation

3 3 Current State of the Art and Benefits of our Approach ● Commercial systems use simple transformation routines – Lack of meta-data limit their extensibility to relevant domains – No mechanism to combine transformations ● Semantic Alignment research attempts to discover mappings between ontologies – The mappings, however, are often expressed in proprietary languages with limited support for inferring additional mappings ● Our approach focuses on how to combine, extend and learn new mappings through inference based on well-supported ontology models and reasoners – Stable under uncertainly and fragmented knowledge – Expressed in OWL, both for high expressivity and for ubiquitous access via the Web – Grows with adoption (follows the open-source model) ● Equally important, our system is designed to work with existing repository technologies – In order to work with operational schemas and data models

4 4 deduced transport-level map Data transformation Semantic Mediation Approach Source_1.XSD Source_2.XSD Source_1.OWLSource_2.OWL data1.xmldata2.xml data1.RDFdata2.RDF Semantic Inference generic code Source_1Source_2 A BC D E F G H L K M = G ◦ L◦F xsd compliance gloze conversionOWL individuals Induced semantic map Legend ● Goal: Transform data1 to data2 ● XSD-compliant XML is a de- facto data transport standard, but XML data are just annotated text with no semantics ● We use the semantics implicit in the XSD by lifting the mediation problem to the OWL/RDF level – Utilizes reasoning tools – Leverages existing knowledge bases of mapping rules – Requires no low-level Java or C++ code, except for generic “glue” code ● Mediator composition is run- time autogenerated (M = G◦L◦F)

5 5 Discovers derived mappings ● Given the following two assertions – Ont1:Terrorist  Ont1:Person (Ont1:Terrorist is a subclass of Ont1:Person). – Ont1:Person  Ont2:Individual, (Person is equivalent to Individual) ● We discover that – Ont1:Terrorist  Ont2:Individual Person Terrorist Individual asserted inferred

6 6 Infers new relationships ● Explicit assertions: – Vessel is equivalent to Ship – hasVector property is equivalent to hasBearing property – The range of hasVector is vector and the range of hasBearing is Bearing ● Through reasoning we can infer that ‘Vector’ is equivalent to ‘Bearing’ (for values of hasVector or hasBearing) – Our Mediation service will not only transform the Vessel information into Ship (asserted mapping), but will also map data elements from Vector to Bearing (inferred mapping) Vessel Ship Location Bearing Vector isEquivalent hasLocation hasVector hasBearing isEquivalent Co-extensive for values of hasVector asserted inferred

7 7 Mediation Framework Ontology A Ontology B AB Mapping Ontology A Data imports B Data imports Mediator Jena in-memory model Pellet imports uses imports persists Jess Other reasoners

8 8 Mediation Algorithm I Ontology AOntology B Type Individual Property communicatesWith Type Person Property isLinkedTo communicatesWith Individual1Individual2 owl:equivalentClass rdfs:subPropertyOf We start by mapping classes and properties in one ontology into those of the other Mapping is expressed in terms of built-in subsumption relations Rules (not shown here) are used to express complex mappings

9 9 Mediation Algorithm II Ontology AOntology B Type Individual Property communicatesWith Type Person Property isLinkedTo communicatesWith Individual1Individual2 owl:equivalentClass rdfs:subPropertyOf Instances in ontology B are created for each instance in A whose type maps to a type in B Properties for the newly created instances in B are asserted whenever their counterparts in A have a mapped property isLinkedTo Person1Person2

10 10 Processing Flow Convert structured and semi- structured data into a common representation Merge data (e.g. merge communication data with profile data) Perform entity resolution (i.e. determine which names refer to the same entities) Semantically enrich data (e.g. create explicit links between persons participating in the same event) Structured Data Transform Merge/Resolve Enrich View Web Data


Download ppt "Inference-based Semantic Mediation and Enrichment for the Semantic Web AAAI SSS-09: Social Semantic Web: Where Web 2.0 Meets Web 3.0 March 25, 2009 Dan."

Similar presentations


Ads by Google