WP2: ONTOLOGY ENRICHMENT METHODOLOGIES Carole Goble (IMG) Robert Stevens (BHIG) Mikel Egaña Aranguren (BHIG) Manchester University Computer Science: IMG: Information Management Group. BHIG: Bio-Health Informatics Group.
INTRODUCTION Current bio-ontologies not very expressive. Ontology enrichment (migration): add richer semantics. ODPs, Normalisation, ULO,... Text mining. Ontology enrichment in CCO.
CURRENT BIO-ONTOLOGIES Difficult for Biologists to exploit expressivity and hence reasoning. Label-centered, not model centered: “positive regulation of ubiquitin ligase activity during meiotic cell cycle” (GO) “acetylcholine biosynthetic process” (GO) Ontologies from text mining or database schemas.
ENRICHMENT From non-expressive to expressive ontologies. Progressive. Already explored implementations: Available in Biological Ontology Next Generation (BONG). Ontology Processing Language (OPL). Based on syntactic/semantic matching. Other implementations in the future: integration of text mining in enrichment.
ENRICHMENT Normalisation. Ontology Design Patterns. Upper Level Ontology. Text mining/learning. Combination of different ontologies.
ONTOLOGY DESIGN PATTERNS Analogous to OOP design patterns: “succesfull modelling recipes”. Abstraction of semantics: better and easier modelling. Documented and repeatable modelling. CCO new possible ODPs: “interaction”, “taxonomy”,...
ONTOLOGY DESIGN PATTERNS Simple Example: Value Partition.
NORMALISATION Hard-coded polyhierarchy: Difficult to maintain: manually add/remove all the relationships. Not expressive: the computer cannot tell why A is a subclass of B.
NORMALISATION Let the reasoner do the job:
SUMMARY - BENNEFITS Tooling. More expressive CCO: Reasoning. Querying. Maintenance. Area not explored in Knowledge Management: publications.