There are a variety of minor discrepancies, errors, and out-of-date in the artifacts in the knowledge base. This item also notes some of them. ontology/term minor refactoring, updated artifacts in KB main
Goal
To get to KB main being 100% usable as is (for mpog trial)
Create a draft release of mpog-pilot v1.3
Make any bug fixes only, mpog-pilot should remain deployable
If main is good to go version it at v1.3, otherwise use mpog-pilot again
To maintain KB main as usable, integrated with CEDAR workflow, PFP development, and MPOG localization
Resolve how to use draft PSDO and CPO terms in cepar for message templates, causal pathways, measures, comparators, preferences
Convert SLOWMO to PFP (project ontology)
Maintain draft PSDP and CPO ontologies in Github (if we resolve the CEDAR generation problem)
Make mpog-pilot branch disappear
But we still need a way release the KB so that we can keep track of which version is used for what
We need to safely evolve the KB in sync with the PFP
Find another way to localize the KB
Approach: make mpog-pilot work, then migrate changes to KB main
Figure out a localization approach. Not this story: See Display-Lab/precision-feedback-pipeline#431
Partial list of items to consider
Some templates and causal pathways may not have been regenerated after changes to the schemas. For example, there are schema:name elements that are not correctly populated
There is a duplication of CEDAR-based "Name" properties and schema.org "schema:name" properties. We should be using schema.org and other standard vocabularies for properties when we can.
Replace app#display-lab in KB (measures)
update KB and PFP to use the current https://schema.org/ namespace (Other tools (like rdflib) expect the current namespace)
Pick a style for properties vs. classes in SLOWMO (see ontology best practices)
In CPs and message templates the "Name" or "name" or "schema:name" properties are conflicting, incorrect, duplicative. Reconcile
performance improving template type is a bad URL; check all these URLs and/or regenerate artifacts from CEDAR
goal worse has no template in cedar so @id is made up
check all@ids in templates and CPs
Check @types. At least MTs have a weird URI
It is possible that we did not update all the DisplayCompatibility to all lower case. Check this
In_to_25% message probably has the wrong @id and other mismatches. It was updated independently of the one on the mpog_pilot branch and we ended up not using it for the trial.
In November some message templates were updated from Cedar and the IDs were changed!! (see this diff). Let us consider maybe using causal pathway instead of template @id if those are going to change. The causal pathways are one-to-one except that ones that use social comparators are three-to-on. BTW, we are not capturing the template @id in the candidate info list.
Motivation
There are a variety of minor discrepancies, errors, and out-of-date in the artifacts in the knowledge base. This item also notes some of them. ontology/term minor refactoring, updated artifacts in KB main
Goal
mpog-pilot
v1.3main
is good to go version it at v1.3, otherwise usempog-pilot
againmpog-pilot
branch disappearApproach: make
mpog-pilot
work, then migrate changes to KB mainFigure out a localization approach. Not this story: See Display-Lab/precision-feedback-pipeline#431
Partial list of items to consider
@id
s in templates and CPs@types
. At least MTs have a weird URIDisplayCompatibility
to all lower case. Check thismpog_pilot
branch and we ended up not using it for the trial.