Dear Philippe, On purpose I provided these examples.
I agree that anatomic structures can be pre-coordinated. They are the exception to the rule. And perhaps in the domain of medical devices we could have exceptions. I see the analogies: - Ontology = Encyclopedia - Terminology = Dictionary - Archetype = Phrase Under specific circumstances we need aggregated concepts for use in the User screen, or during data entry, or data export (for statistics) SNOMED has the capability to create those complex codes for these very specific purposes. When we need to store and retrieve in a database or when a Clinical Reasoner processes data the Rules apply. Archetypes help define the datum but also its full context/epistomology so the data can be interpreted safely. Gerard Freriks +31 620347088 [email protected] Kattensingel 20 2801 CA Gouda the Netherlands > On 28 Mar 2018, at 18:14, Philippe Ameline <[email protected]> wrote: > > Gerard, I like your rule (build a grammar that coordinates a vocabulary of > "atomic concepts" instead of agglutinating meta-concepts in the vocabulary), > but not your "left eye" example ;-) > > In my own "universe", the "left eye" is a true physical object (when "eye" is > a concept). > I would say the same thing about the "left common carotid artery" which is > something you can actually "touch". Representing this concept as an artery > which belongs to the carotid network on the left part of the body is not > pre-coordination but rather what could be expressed by a semantic network. > > I agree with your "blue eye" example since it is simply "an eye which color > is blue". > > To answer Mikael Nyström in the same message, I would be OK with his > conclusion about "the simple solution is to not use what you don't need" if : > - this task had not been unsuccessfully tested by Belgian GPs, who literally > spent years trying to select a subset with no avail, > - this "language" was not advertised as a communication system... which > actually keeps you exposed to the full set unless everyone agrees to a common > subset (using Tom's proposals and Gerard's rule, for example). > > Mikael is right when he says that it would be unfair to say that Snomed "is > bad because it isn’t optimized for [one's] narrow use case"... unless there > are lots of narrows use cases needed and unless one of these narrow use cases > is key to the next mainstream use case. If innovators (typically the ones > with weird use cases) consider Snomed as a millstone around their necks, it > is actually an issue worth considering. > As pointed before in this thread, this conversation is closely related to HL7 > V3 and its RIM before it was "fhired". > > Best, > > Philippe
signature.asc
Description: Message signed with OpenPGP
_______________________________________________ openEHR-technical mailing list [email protected] http://lists.openehr.org/mailman/listinfo/openehr-technical_lists.openehr.org
