Modularisation of SNOMED CT

This is to discuss the proposal for the modularisation of SNOMED CT.

As discussed in the face to face meeting in Montevideo, here is a link to a document I (@Dion McMurtrie) wrote last year called Modularised SNOMED CT

This document was written as an exploration of how to make SNOMED CT less monolithic, and what the benefits and issues might be. Thinking like a software developer it seems appealing split SNOMED CT into more modular pieces, much like software is.

I'd thought about this for a while but chose to write it up and make it a bit more real when considering the Palo Alto meeting on SNOMED CT's DL feature set and possible OWL renderings. It feels appealing for SNOMED CT modules to be rendered as useful OWL modules with import declarations matching the Module Dependency Reference Set.

Interestingly it is also possible to derive modules from a monolithic ontology as well, for example refer to ontology modularity research at the University of Manchester. Their approach is computing modules rather than dividing the authoring (perhaps because they can't control the source?), but clearly see value in a modularised ontology. One statement made on their website is

...SNOMED’s sheer size makes it very difficult to use, maintain, browse, understand, debug and reuse it, to derive and explain inferences. It is therefore a highly desired task to generate topic-specific sub-dictionaries on pressing a button. For example, if only the knowledge about skeletal bones is of interest, users need to be able to work with the corresponding subontology. If that subontology encapsulates all knowledge that is relevant for the bones taxonomy by including knowledge about, say, diseases affecting bones, then it is called a module. For OWL ontologies, the task of extracting a module for import or reuse is well-understood.

From the Montevideo meeting it seemed clear that at least breaking "core" from "non-core" (extension/derivative work like the ICD-10 map) content in the International release of SNOMED CT would be good. This would enable organisations/nations to optionally take the "non-core" content and aren't forced to republish it with their extensions.

However, it would be good to consider what opportunities further modularisation may provide and possible approaches, and to that end it would be good to hear comments.

Please use the comment feature at the bottom of the page to raise questions, issues or make comments. Feel free to comment in the Google Doc as well, but please post major thoughts on discussing this topic back on this page.