added UCxn annotation#551
Conversation
|
Thanks! One problem: this is overriding existing Cxn annotations (for relative clauses, which start with |
|
That's the way that @s-herrera 's script currently works, I'm not sure if it needs to stay that way or could be changed |
| 4 Bay Bay PROPN NNP Number=Sing 5 compound 5:compound _ | ||
| 5 View View PROPN NNP Number=Sing 3 obj 3:obj _ | ||
| 6 if if SCONJ IN _ 9 mark 9:mark _ | ||
| 7 you you PRON PRP Case=Nom|Person=2|PronType=Prs 9 nsubj 9:nsubj _ | ||
| 8 are be AUX VBP Mood=Ind|Number=Sing|Person=2|Tense=Pres|VerbForm=Fin 9 aux 9:aux _ | ||
| 9 looking look VERB VBG Tense=Pres|VerbForm=Part 3 advcl 3:advcl:if _ | ||
| 9 looking look VERB VBG Tense=Pres|VerbForm=Part 3 advcl 3:advcl:if CxnElt=3:Conditional-UnspecifiedEpistemic-NoInversion.Protasis |
There was a problem hiding this comment.
Happened to notice that this one is a speech act conditional. I don't think we have a specific label for those.
There was a problem hiding this comment.
Happened to notice that this one is a speech act conditional. I don't think we have a specific label for those.
Hi, @nschneid !
I guess it wouldn't be right to tag «if» in features as a Protasis marker, but it would be helpful for distinguishing the «conditional» marking in Mordvin languages, where the verbs take a protasis morpheme and called «Conditional» in traditional grammars.
Would you have any suggestions for a «Protasis» feature?
There was a problem hiding this comment.
@rueter The UCxn approach implemented here is to identify the content elements of the construction via Grew rules that incorporate grammatical marking. We don't tag the functional morphemes like "if" directly. UD morphological features (at least as I understand them) are generally for indicating where a word form lies in a paradigm, not for what an uninflecting function word like "if" contributes to an analytic construction. Does that answer your question?
Yes, it removes all the existing Cxn annotations. It can be changed, but it means that two different annotations for the Cxn field will coexist in the same treebank. |
|
I think we need to be able to support Cxn annotations that are not from the Grew rules, because many treebank maintainers will want to add language-specific constructions by other means (other automatic means, manual annotation, or a combination thereof). |
|
Yes, I agree. I was just saying that, in this particular case, two naming convention will coexist in the same treebank. I have sent the new treebanks to @LeonieWeissweiler so she can update the PR. |
|
PR has been updated! |
| 11 how how ADV WRB PronType=Int 12 advmod 12:advmod _ | ||
| 12 fanatic fanatic ADJ JJ Degree=Pos 9 ccomp 9:ccomp _ | ||
| 11 how how ADV WRB PronType=Int 12 advmod 12:advmod CxnElt=12:Interrogative-WHInfo-Indirect.WHWord | ||
| 12 fanatic fanatic ADJ JJ Degree=Pos 9 ccomp 9:ccomp Cxn=Interrogative-WHInfo-Indirect|CxnElt=12:Interrogative-WHInfo-Indirect.Clause |
There was a problem hiding this comment.
I think this is actually exclamative not interrogative (but it would require manual disambiguation).
| 12 they they PRON PRP Case=Nom|Number=Plur|Person=3|PronType=Prs 13 nsubj 13:nsubj _ | ||
| 13 call call VERB VBP Mood=Ind|Number=Plur|Person=3|Tense=Pres|VerbForm=Fin 10 ccomp 10:ccomp _ | ||
| 13 call call VERB VBP Mood=Ind|Number=Plur|Person=3|Tense=Pres|VerbForm=Fin 10 ccomp 10:ccomp Cxn=Interrogative-WHInfo-Indirect|CxnElt=13:Interrogative-WHInfo-Indirect.Clause |
There was a problem hiding this comment.
Tree structure is wrong—should actually be a free relative, #278 (there are a lot of these errors unfortunately, just noting for later)
|
Did some spot checks. @LeonieWeissweiler @s-herrera would you mind fixing the Two limitations that I noticed for documenting in the README:
|
|
the wordform empty issue has been fixed, thanks for pointing that out! |
|
Wes has added a rule to fix the polar interrogative existential issue. Sorry to bug you but would you mind running the rules again? |
|
fixed now, all credit to @s-herrera who keeps running them :) |
|
Looking at the last commit, something went wrong...maybe it was based on an earlier version of the data? |
|
Yes, sorry about that... I fixed it and sent, hopefully, the corrected files to @LeonieWeissweiler. |
|
Great. I undid the spurious commit. |
|
does this one work? |
|
I suspect the files are correct, but so we can see the changes, the new commit should be based on the "removed wordform empty" commit not the buggy "fixes to UCxn polarity questions" commit. I will attempt a rebase. |
…rporate rule update from LeonieWeissweiler/UCxn@140ad7b
| 4 not not PART RB Polarity=Neg 5 advmod 5:advmod _ | ||
| 5 available available ADJ JJ Degree=Pos 6 advcl 6:advcl:if CxnElt=6:Conditional-UnspecifiedEpistemic-NoInversion.Protasis | ||
| 6 is be VERB VBZ Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin 0 root 0:root Cxn=Conditional-UnspecifiedEpistemic-NoInversion,Existential-CopPred-ThereExpl|CxnElt=6:Conditional-UnspecifiedEpistemic-NoInversion.Apodosis | ||
| 6 is be VERB VBZ Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin 0 root 0:root Cxn=Conditional-UnspecifiedEpistemic-NoInversion,Existential-CopPred-ThereExpl,Interrogative-Polar-Direct|CxnElt=6:Conditional-UnspecifiedEpistemic-NoInversion.Apodosis,6:Interrogative-Polar-Direct.Clause |
There was a problem hiding this comment.
Good example of a sentence with 3 different kinds of constructions
|
I learned a new trick: |
@rueter if you're looking to annotate words like "if" in conditionals, that sounds a bit like discourse relation + discourse marker annotation (in eRST) or connective annotation (in PDTB) - UD_English-GUM annotates those in both frameworks, so you can take a look there/let me know if you want more information on either. This includes disambiguating the relation signaled by e.g. "if", which is not always a conditional (for example: "there's water in the fridge if you want some"). |
|
Merged with a README note, thanks all! |
Adds construction annotation in the MISC column, as described in https://aclanthology.org/2024.lrec-main.1471.pdf
Annotates NPN, Interrogatives, Conditionals, and Existentials, on the head of the respective phrase, plus construction elements.
For full specification and a subset of the treebank with only cxn-annotated sentences see here: https://github.com/LeonieWeissweiler/UCxn