Add instance level annotations capability to RM

Description

This is a requirement for semantic annotation of Archetyped data for use in semantic web applications. Onological annotation is an important mechanism for discovery and automated inferencing that is commonly used in research. One practical example would be to add Gene and Protein sequence URIs to data items (CLUSTER or ELEMENT) or higher up the Archetype structure based on Instance data, for example in the case of certain lab results corresponding UniProt database URI can be saved together with instance data. This would result in consistent annotation and semantic querying of Archetype based repositories using general purpose semantic web tools. It would be desirable to follow RDF approach; e.g.
subject: unique runtime archetype path,
predicate: URI (alternatively can be DV_CODED__TEXT or some openEHR enumeration such as MIRIAM annotation types),
object: URI or CODE_PHRASE (that will have external terminology link)

See: MIRIAM (Minimum Information Required In The Annotation of Models): http://en.wikipedia.org/wiki/Minimum_information_required_in_the_annotation_of_models

Environment

None

Activity

Show:
Koray Atalag
May 7, 2015, 11:36 PM
Thomas Beale
May 8, 2015, 12:45 PM

The challenge here is that such annotations would need to be added before commit, so that versioning semantics are preserved. If people want to go adding this kind of thing after the fact, which seems likely, then we should not put this kind of annotation in the EHR data (which is after all, point of care data), it should go in a separate index that correlates EHR URIs (with full paths to data items) and one or more ontological markers.

This would require a new piece of RM, and a new service, which I think would be appropriate things to do.

Koray Atalag
May 14, 2015, 5:27 AM

I see your point and I am aware this needs to be depicted during commit - that is entirely possible. Besides current DV_TEXT package already has TERM_MAPPING class which I think serves for exactly the same kind of purpose - it is just not compatible with modern semantic annotation formalisms (mostly RDF based these days). Probably we need to revisit this class.

Thanks if you can elaborate on the approach you are suggesting - e.g. new RM, service etc. Is this something you suggest taking to the meeting? This is pretty foundational for my research.

Reporter

Koray Atalag

Labels

None

Affects versions

Priority

Minor
Configure