160 likes | 169 Views
This paper discusses the process of building corporate knowledge by integrating ontologies, providing a mathematical framework for conceptual graphs, and demonstrating an improved ontology merging algorithm.
E N D
Building Corporate Knowledge through Ontology Integration Philip Nguyen PhD, Principal Technical Specialist, Department of Justice State Government of South Australia - Australia Dan Corbett PhD, Principal Scientist, Science Applications International Corporation Washington DC, USA PKAW’06, Guilin, China, 7-8 August 2006
Corporate Knowledge (source: Frappaolo, C., Wilson, L., “After the Gold Rush - Harvesting Corporate Knowledge Resources”, Intelligent Knowledge Management, 2001) http://www.gse.harvard.edu/~t656_web/Spring_2002_students/kothuri_smita_knowledge_in_orgs.htm
Upper Ontology Task Ontology Domain Ontology Application Ontology Ontology Definition • Tom Gruber: Ontology = a specification of a conceptualization • Nicola Guarino: Ontology Categories
Our Ontology Formalization Canon Ontology 1 Ontology 2 Ontology 3 Ontology = semantically consistent subset of a Canon Canon = Universal Ontology
Real World Abstract World I TR TC B conf Canon / Ontology K = (T, I,, conf, B) P. Nguyen and D. Corbett. "A Basic Mathematical Framework for Conceptual Graphs," IEEE Transactions on Knowledge and Data Engineering, vol. 18, no. 2, pp. 261-271, February, 2006
Concepts & Relations K = (T, I,, conf, B) • T: set of concepts (TC ) & relations (TR ) • : subsumption relation on concepts & relations • Concepts: MalePersonPersonLivingEntity • Relations (or Relational Concepts): isChild (YoungPerson, FemalePerson, MalePerson) isSon (YoungMalePerson, FemalePerson, MalePerson) isSon isChild • Future ISO Standard on Common Logic: http://cl.tamu.edu/
Individuals & Type Conformance K = (T, I,, conf, B) • I: set of individuals • conf: conformance relation conf : I TC e.g. conf (“Peter”) = [Man] Peter is a man, a person and a living being, i.e. Man = infimum (person, living being, …)
Relations & Canonical Formation K = (T, I,, conf, B) B: TR {TC } B (relation) = ordered list of concepts e.g. B (“isChild”) = {YoungPerson, FemalePerson, MalePerson} B (“isSon”) = {YoungMalePerson, FemalePerson, MalePerson} isSon isChild semantic inclusion & YoungMalePerson YoungPerson FemalePerson FemalePerson MalePerson MalePerson
3 Ontology Integration / Merging Concepts & Relations missing from canon Expert Validation Canon 2 Expert Decisions 1 MultiOntoMerge Algorithm Source Ontologies Merged Ontology 4 Improved Canon
Ontology Merging - Example Animal Beast Fish-eater Bird Fish-eater Cormorant Input Ontology 2 Galah Pelican Input Ontology 1
Animal Animal Bird Fish-eater Bird Fish-eater Galah Pelican Galah Pelican Cormorant Input Ontology 1 Merged Ontology – Step 1 (after Semantic Compaction) Ontology Merging – Step 1 Beast Fish-eater Cormorant Input Ontology 2
Animal Animal Bird Fish-eater Bird Fish-eater Galah Pelican Cormorant Galah Pelican Cormorant Merged Ontology – Step 1 (after Semantic Compaction) Merged Ontology – Step 2 (after Semantic Completion) Ontology Merging – Step 2
Ontology Merging – Step 3 Animal Animal Bird Fish-eater Bird Fish-eater Galah Fish-eater-bird (*) Galah Pelican Cormorant Merged Ontology – Step 2 (after Semantic Completion) Pelican Cormorant Final Merged Ontology – Step 3 (after ensuring Mathematical Soundness) (*) osprey
3 Ontology Integration / Merging Concepts & Relations missing from canon Expert Validation Canon 2 Expert Decisions 1 MultiOntoMerge Algorithm Source Ontologies Merged Ontology 4 Improved Canon
Text & Graph Representations • animal > bird, fish-eater • bird > fish-eater-bird, galah • - fish-eater > fish-eater-bird • - fish-eater-bird > cormorant, pelican http://users.on.net/~pnguyen/
Conclusion • Corporate Knowledge = Canon • Canon built from successive merging of existing domain ontologies • Each ontology merging exercise more automated than previous ones • Canon is at all times semantically compact, semantically complete and mathematically sound