SNOMED CT OWL2 EL Logic Profile
Description
The purpose of this work item is to develop normative guidance on representing and maintaining SNOMED as an OWL Ontology. This involves listing the logic constructs in the EL2 profile that would be included in SNOMED and those that would not, providing a rational for the inclusion or exclusion of each construct.
This work draws on the work of a SCT/DL Subgroup, which met in Palo Alto in December 2014. Their paper can be found here. The normative guidance would be presented in two sections. The first section would repackage parts of the Palo Alto paper relating to the inclusion and exclusion of logic constructs. The second section assumes that SNOMED is a specialization of OWL, and explains how the included constructs would be used to meet SNOMED specific requirements.
Objectives
To develop a draft standards document that sets out the proposed OWL 2 EL logic profile for use in SNOMED CT.
Which DL constructs are to be used with SNOMED, and which are not, and why.
To which parts of the SNOMED concept model each logic construct applies.
Limitations on the use of each construct/concept model application. Assumes that the OWL 2 EL constructs would be restricted for use in SNOMED.
Initially the audience for the document would be the IHTSDO Management Team, the broader Modeling Advisory group, and possibly the Editorial Advisory Group. We would incorporate feedback from those groups into the document and put it out for community review.
IHTSDO would provide the project management structure for the standards process; this group would focus on the content of the document.
Note that editorial guidance is required but would be an objective of a separate document.
Consultation
A publicly visible consultation page has been posted at SNOMED CT Description Logic Enhancements - Consultation
Press Release and Proposal
A "Press Release" for this work has been issued : SNOMED CT Logic Profile Enhancement
Which links to the main proposal document: https://docs.google.com/document/d/1tqNEA6S4fEF4fgj15OPabYA2E0VTz8epxvRRwczKizQ
Next Sub Group Meeting
Tuesday 6 - Thursday 8 June 2017 08:30 UTC - 16:00 UTC
https://snomed.zoom.us/j/535528933
Meeting Key Points
Thursday 9 February
Suggestion that our 'degraded position' from OWL to RF2 would be to take the inferred relationships created when classifying the OWL and make these the stated form in RF2. My understanding [ PWI ] is that these would need to then be primitive since reclassifying the inferred view without all the axioms in it would give different results.
Defining concepts using information that is not presented in the RF2 Relationship files is incompatible with SNOMED Expressions and Expression Constraint Language.
Examples are needed for further discussion.
Thursday 23 February
Suggestion that a distribution file for OWL could use 1 row per axiom, rather than version at the concept level.
Discussion on issues with ECL and multiple sufficient sets.
Discussion that the inferred form should describe just the necessary relationships - or that we should flag which relationships are necessary in the RF2.
Wednesday 5 April - Key topics of concern & interest:
PWI: Impact of stating axioms using OWL on SNOMED Family of Languages (Michael L?) DK: Especially ECL.
PWI: What can be represented in inferred relationship file? For example, concrete domains: how would a user consume those. Would they have to classify the OWL file to get those attribute values? DK: The impact on existing implementations.
HS: Defined DL expressivity. Will need a way of stating disjointness (things which can't simultaneously exist). Distribution method. People use hierarchical relationships...even moving into "Role Groups" territory can be a challenge.
ML: Full range of OWL2 EL available for modeling. Impact a concern, dealing with communication and "selling" of proposal. ECL & structural subsumption - touches on defn of various normal forms. Keeping in mind users who use SNOMED in tabular form. Arbitrary nesting will cause problems for them - usability challenge. Range of audience. SNOMED currently in sweet spot of simplified EL play off between power of expressivity and simplicity.
Negation should be kept out of scope. Too large a change, too complex - would prevent completion ie keeping it to EL. "A is not B is different from A and B are disjoint" PH: "If A and B are disjoint, then A is not B" HS: "Could be another thing entirely" ML: "Reasoning from the negated concept that you can't do with disjointness." Disjointness allows for the discovery of inconsistencies. (Harold thinks we will turn up a number of inconsistencies in SNOMED if/when this functionality is added).
DK: Disjointness - SNOMED top level hierarchies might not be disjoint (substances which are products).
PH: Pain point of moving users off RF2. Adding dis-jointness (brings in negation and all (not the case, discuss offline)). Won't be able to walk away from RF2/ECL, so neither can we have pure OWL DL. We will need both OWL and RF2. Kaiser represented additional inferred relationships as if a user "Dr Reasoner states..."
ML : Modularisation of the OWL file - sections pulled in by import statements, divided up by existing moduleids.
DK: People want different (OWL) languages - new PERL script? HS: Mayo clinic script - political issues? Discuss again in London. Has comparison utility. ML : ADHA also have their own variation of the Spackman transformation (possibly exists on a branch of same) is module aware + supports concrete domains in refsets if present.
Status | IN Progress |
|---|
Team
@Former user (Deleted)
@Daniel Karlsson (Unlicensed)
@Former user (Deleted)
@Peter Williams
@Ronald Cornet (Unlicensed)
@Yongsheng Gao
@Dion McMurtrie
@michael lawley
@Guillermo Reynoso
Child Pages
Related: Naming convention of new OWL refset file
Previous Sub Group Meetings
Copyright © 2025, SNOMED International