Open rhdunn opened 11 months ago
I could see using Ord for the numerical ones, but until we sort out what we're doing about LS I will leave this open. I anticipate this will stay as-is for v2.13.
Ping regarding this ( and @nschneid) ... one of the more frequent errors caused by the CoreNLP constituency -> dependency converter is because it wants to make the dependency "num" but the UPOS "X". If we come up with a standard and apply it to the EWT & GUM treebanks, I can implement that in the converter pretty easily.
Yeah, we need a standard. It's under discussion in the core group.
X
UPOS. -- EWT favoursNUM
for these and https://universaldependencies.org/u/pos/X.html states that it should be used restrictively.NumType=Ord
feature (as they specify an ordered list of items).1.
etc variants should have theNumForm=Digit
feature.a)
etc variants should have aNumForm
feature, but no suitable form currently exists for these; maybeNumForm=Alpha
(alphabetic -- "Examples: a, b, c, α, β, γ").(
and)
as separate tokens.Validation issues: