Suspicious cc attachment

[nschneid]

Ordinarily the cc marks a conjunct attaching via conj, or it may be sentence-initial (in which case the head probably attaches to root). But not always:

Queries:

• GUM
• EWT

If there is a coordinating conjunction, I assume we should prefer a conj analysis over alternatives like parataxis or advcl? Some patterns that show up as non-conj:

• 10 days or so
• SENTENCE, and so on.
• SENTENCE, but not when...
• SENTENCE, but only after V-ing.

https://universaldependencies.org/u/dep/parataxis.html mentions "non-conjunction connective[s]", which suggests that clauses connected by conjunctions are conj instead. Also, conj seems to be in the majority when there is a parenthesis: https://universal.grew.fr/?custom=675a54dfb871c

[nschneid]

• Constraints on cc parent function implemented in GUM validator amir-zeldes/gum@efab974 (and I will add to neaten.py)

@amir-zeldes do you agree with the above that conj should be preferred for "and so on" etc.?

[amir-zeldes]

Yes, definitely!

[nschneid]

Which should be the head of "so on"—"so" or "on"?