Snorocket 2.0: Concrete Domains and Concurrent Classication

of 12/12
Snorocket 2.0: Concrete Domains and Concurrent Classication Alejandro Metke-Jimenez | Postdoc, AEHRC [email protected] THE AUSTRALIAN E-HEALTH RESEARCH CENTRE | ICT CENTRE
  • date post

    24-Feb-2016
  • Category

    Documents

  • view

    34
  • download

    0

Embed Size (px)

description

Snorocket 2.0: Concrete Domains and Concurrent Classication. the Australian e-Health research centre | ict centre. Alejandro Metke -Jimenez | Postdoc, AEHRC . [email protected] csiro.au. Snorocket 2.0. Motivation Architecture Concrete domains Test ontologies Experimental results - PowerPoint PPT Presentation

Transcript of Snorocket 2.0: Concrete Domains and Concurrent Classication

Digital Productivity Portfolio

Snorocket 2.0: Concrete Domains andConcurrent ClassicationAlejandro Metke-Jimenez | Postdoc, AEHRC [email protected] Australian e-Health research centre | ict centre

1Snorocket 2.0MotivationArchitectureConcrete domainsTest ontologiesExperimental resultsFuture workOur ontology tools

2MotivationSnorocket was initially implemented to support SNOMED CT authoringSome concepts cannot be correctly modelled without concrete domainsFor example, Hydrochlorothiazide 50mg tablet cannot be modelled without a data literal to represent the quantity of the active ingredientAMTv3 has recently introduced concrete domains3Architecture

4Architecture

5Concrete DomainsCEL normalisation algorithm works with minimal changesAdded an additional completion rule

6Test OntologiesSNOMED CTAustralian Medicines Terminology v3 (with concrete domains)AMT axioms are more complex than the typical axioms found in SNOMED CTHigh degree of nesting (e.g. to associate BOSS value and unit with actual ingredient)Still simple to author since axioms follow same patternOntology# Classes# Object Properties# Data Properties# AxiomsOriginalNormalisedSNOMED CT2965186206606101169613AMTv3610597841507505613317Test Ontologies

8Experimental ResultsPerformance of tableaux-based reasoners was poor when classifying AMTElk is fastestSNOMED CTAMTHermiT (1.3.7)1567.3memFACT++ (1.6.2)3304220jCel (0.15)761-Snorocket (1.3.4)33.8-Snorocket (2.0.1)2626.2ELK (0.32)9.110.59Future workOptimise multi-threaded implementationIncorporate restrictions necessary to ensure tractability when dealing with concrete domainsExplore mechanisms to incorporate units natively in reasoning process

10Our ontology toolsSnorocket ontology reasonerOpen source Apache 2.0Available in GitHub: https://github.com/aehrc/snorocketOntoserver ontology searchProvides search over ontologiesAvailable as a Java library and a REST APIsandbox: http://ontoserver.csiro.au:8080Based on Lucene and SolrSnapperEclipse-based SNOMED CT mapping & refset toolFree for use in Australia11Questions?Australian eHealth Research CentreAlejandro MetkePostdoctoral Research Fellow

Phone: 07 3253 3645Email: [email protected]: www.aehrc.com

12