Authors declare a label for a node, which we generally do not wish to alter. Generally speaking, the top search hit from the grounding-search is associated to that node/label. A user can select re-assign the ground by selecting another record from the search hits. But what happens when the database record you want isn’t in those search hits? The desire, in this case, it to directly associate a node/label in the UI.
This would mainly impact re-curation by expert curators or biofactoid team members who wish to correct obvious errors.
There are several instances where this arises:
Synonym
Label isn’t indexed as a name or synonym in grounding-search
An indirect solution we have is to supplement the grounding-search record with a synonym via a “patch” file.
e.g. Label “CARP-1” isn’t listed as a synonym of “CCAR1” (NCBI Gene 55479)
e.g. Label “Telomerase” isn’t listed (?) as a synonym for “Tert” (NCBI Gene 21752)
Modifier
Prefixes often skew the search results: e.g. mtRNA, e.g. OGT Human
Near match
e.g. label “MT2_Mm” could reasonably be associated with the closest match “MERV-L” (NCBI Gene 27282)
Out-of-Dictionary
Allow null association
It would be nice if the grounding-search could detect poor matches, but this is not yet implemented.
It is better to assign NO association than to leave an erroneous association in place
Other: Intriguing possibility is to expand the range of concepts supported without compromising search by allowing GET of many more resources than are available via SEARCH.
e.g. Species: Could allow GET on entities from other/all organisms in NCBI Gene (e.g. Gallus Gallus)
Authors declare a label for a node, which we generally do not wish to alter. Generally speaking, the top search hit from the grounding-search is associated to that node/label. A user can select re-assign the ground by selecting another record from the search hits. But what happens when the database record you want isn’t in those search hits? The desire, in this case, it to directly associate a node/label in the UI.
This would mainly impact re-curation by expert curators or biofactoid team members who wish to correct obvious errors.
There are several instances where this arises: