Implement nmod:desc for honorific pre-nominal titles: Mr., Dr., etc.

[nschneid]

• EWT abbreviated titles and their expanded equivalents: https://universal.grew.fr/?custom=6769d90c72a6e https://universal.grew.fr/?custom=676acf880dbd0

Manually implementing the few that are not flat (many instances of "St." for "Saint" are compound). Will make a rule for the flat ones.

[nschneid]

• EWT King, Queen, etc. as compound

[nschneid]

I reviewed flat-initial PROPNs in GUM and EWT and filtered out the honorifics. The lexical lists in the query can be used for a conversion rule.

In the above review, found a few GUM errors:

• PROPN should be ADJ: Emeritus (Professor), Golden (Gate), Chief (Justice)
• wrong structure (I would go with compound): Comic Corps, Rationalist Atheists, Senate Democrats, State Department, NASA Administrator, US President, Vice President

N.B. "Dean" and "Earl" are in principle ambiguous between titles and first names, but in the data they are only the latter.

[nschneid]

Mostly done with EWT but a few edeps from relative clauses to relativizer antecedents got deleted—need to investigate. (I think 2 new rules after invert headedness will solve this: enhanced relation constraints #4~#1;#2.*#4, #4~#1;#4.*#2, mark #4:edep as OLD and add #4~#2. Where #1 is the honorific, now the head of #2, which begins the name part.)

[amir-zeldes]

Could you share your whole depedit script as a starting point for GUM too? We won't need the edeps part but the basic dependencies would benefit from it!

[AngledLuffa]

Oh, if such a script already exists, we could run that on ParTUT as well. Could pass it along to @LarsAhrenberg for LinES

is updating ParTUT going to be a community effort? seems like the author is still out there occasionally doing things, but only very occasionally

[nschneid]

https://github.com/UniversalDependencies/UD_English-EWT/blob/64538925ef30bf5b4af70039aff564afc8608159/not-to-release/tools/flat2nmoddesc.ini

But see the note at the end about nested cases requiring manual attention.

[nschneid]

is updating ParTUT going to be a community effort?

I can answer annotation questions about other treebanks but in terms of proactive updates, I have my hands pretty well full maintaining EWT.

[amir-zeldes]

Yeah, same here - I have my hands full at the moment I'm afraid!

[AngledLuffa]

phrases such as The playwright and critic George Bernard Shaw do not get updated with nmod:desc?

and what is the correct arrangement for Drs. Foo and Bar? nmod:desc(Foo, Drs.) and maybe an enhanced dependency nmod:desc(Bar, Drs.) if the treebank supports it, which is not the case for ParTUT anyway?

[nschneid]

phrases such as The playwright and critic George Bernard Shaw do not get updated with nmod:desc?

No, the definite article indicates that there are 2 full nominal phrases (related by appos)

and what is the correct arrangement for Drs. Foo and Bar? nmod:desc(Foo, Drs.)

Yes

and maybe an enhanced dependency nmod:desc(Bar, Drs.) if the treebank supports it, which is not the case for ParTUT anyway?

Not sure about the enhanced dependency, especially since "Drs." is explicitly plural, and thus applies to the full coordination

[amir-zeldes]

So, quick question - do we use nmod:desc for "St. Petersburg"?

[nschneid]

I'd say yes, because the contribution of the "St." originally is as a title (of a name that got turned into a place name)

[amir-zeldes]

👍 Doing the same with St. Louis