How to determine if a verb form is treated as NOUN or should also have the VerbForm=Vnoun feature?

[AngledLuffa]

So in English, there are a few ways of turning a verb into a noun:

# writer
1       This    this    DET     DT      Number=Sing|PronType=Dem        2       det     2:det   _
2       writer  writer  NOUN    NN      Number=Sing     6       nsubj   6:nsubj _
3       has     have    AUX     VBZ     Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin   6       aux     6:aux   _
4       for     for     ADP     IN      _       5       case    5:case  _
5       years   year    NOUN    NNS     Number=Plur     6       obl     6:obl:for       _
...

# writing
# text = I sure like his style of writing.
1       I       I       PRON    PRP     Case=Nom|Number=Sing|Person=1|PronType=Prs      3       nsubj   3:nsubj _
2       sure    sure    ADV     RB      _       3       advmod  3:advmod        _
3       like    like    VERB    VBP     Mood=Ind|Number=Sing|Person=1|Tense=Pres|VerbForm=Fin   0       root    0:root  _
4       his     his     PRON    PRP$    Case=Gen|Gender=Masc|Number=Sing|Person=3|Poss=Yes|PronType=Prs 5       nmod:poss       5:nmod:poss     _
5       style   style   NOUN    NN      Number=Sing     3       obj     3:obj   _
6       of      of      ADP     IN      _       7       case    7:case  _
7       writing writing NOUN    NN      Number=Sing     5       nmod    5:nmod:of       SpaceAfter=No
8       .       .       PUNCT   .       _       3       punct   3:punct _

There's also other examples of nouns derived from verbs on Wikipedia:

https://en.wikipedia.org/wiki/Verbal_noun

such as discovery:

1       The     the     DET     DT      Definite=Def|PronType=Art       2       det     2:det   _
2       discovery       discovery       NOUN    NN      Number=Sing     0       root    0:root  _
3       of      of      ADP     IN      _       5       case    5:case  _
4       a       a       DET     DT      Definite=Ind|PronType=Art       5       det     5:det   _
5       number  number  NOUN    NN      Number=Sing     2       nmod    2:nmod:of       _
6       of      of      ADP     IN      _       8       case    8:case  _
7       '       '       PUNCT   ``      _       8       punct   8:punct SpaceAfter=No
8       bugs    bug     NOUN    NNS     Number=Plur     5       nmod    5:nmod:of       SpaceAfter=No
9       '       '       PUNCT   ''      _       8       punct   8:punct _

None of these are treated as Vnoun in the annotations.

I'm wondering, at what point do we decide to start annotating Vnoun for a new annotation scheme? What would justify doing so, and in what situations would one prefer to use NOUN with some other featurization? (neither writer nor writing have any features indicating their derivation from write in the above examples, nor does discovery)

[nschneid]

"Verbal noun" is a grammatical term used for some languages, where there are categories of words that are grammatically sort of like verbs but are not gerunds or infinitives. The term is not used for English as far as I'm aware.

("Discovery" is a deverbal noun—morphologically it has a verb root but the full word displays no properties of verbs.)

[AngledLuffa]

The term is not used for English as far as I'm aware.

But... why not? What grammatical features make it more necessary to annotate as a Vnoun instead of just another NOUN class?

[nschneid]

Well, the documentation at https://universaldependencies.org/u/feat/VerbForm.html#ger-gerund suggests that "verbal noun" is a better term than "gerund", but the latter is what we have been using for English (and the term I typically hear in reference to English or Latin). I defer to @dan-zeman on this one.

[AngledLuffa]

Would love to hear thoughts from @dan-zeman or anyone else who has used (or deliberately not used) Vnoun. Thanks!

[dan-zeman]

VerbForm=Vnoun is for forms that are morphologically noun-like, derived from the verb stem, but they still refer to the action, not one of the participants. So drinking would fit, but drinker or drink would not. The latter could refer either to the substance in your glass, or to the action; but I still would not use VerbForm=Vnoun for it. We have been using it for rather productive morphology that uses specific affixes, not just any derivation. Specifically, in English there seem to be hundreds of stems that can be treated either as nouns or as verbs; but I don't think we need to mark the nouns in this category with the feature.

In Czech, we have a very productive suffix -ní/-tí which, when added to a verbal stem, gives you a noun (and can then further inflect as nouns, e.g. for Case, which most verb forms cannot do). Other Slavic languages have similar morphology, too. I use VerbForm=Vnoun to label forms that have that suffix. For example, obchodovat "to (do) trade" (VerbForm=Inf) > obchodování "trading" (VerbForm=Vnoun). In contrast, there is also the noun obchod "trade" but I do not label it with any value of VerbForm, although it usually refers to an event (but it can also refer to a store, i.e., to a place where trading is done). An open question is what to do when the productively created verbal noun starts to be used in a shifted meaning, e.g., pití "drinking" may still refer to the action, but in other contexts it may refer to the liquids that will be consumed. I actually do not know what happens in the Czech data with these; I would have to invest some time into investigation.

[Stormur]

Dan has already explained the issue.

I hope I am forgiven If I self-cite my article where I also touch upon verbal nouns (in fact what sparked it was a study on Latin so-called gerund/ive).

Terminologically speaking, grammars give many names to similar nominal forms of verb, but for sure "verbal noun" or Vnoun should be the most neutral of all, irrespective of the tradition that most uses it. In a couple of past releases (at least in the dev stage) Latin had Vnoun for what is now labelled as Inf, for the only reason that that change in denomination sparked furious and irreconcilable debates in the annotating community. But the fact stands that what is called infinitive is the most standard verbal noun that can be imagined.

I would say that a verbal noun, to still be a verb form, has to show a "verbal pattern". So, in the examples that you give, writing could take an object exactly as any so-called finite counterpart of its (and this is also how the Latin infinitive works), but writer cannot (as has been noted, writer is deverbal, not verbal).

As usual, once a form has been identified as a verbal noun it should always be annotated as VERB and not oscillate between, say, VERB and NOUN (now, incidentally, in English things look more complicated as there seem to be two formally convergent forms in -ing...).

[amir-zeldes]

We have been using it for rather productive morphology

This is the main point for my understanding of this annotation in UD - for the feature to apply, it has to be a form of productive, automatically applicable morphology

I would say that a verbal noun, to still be a verb form, has to show a "verbal pattern". So, in the examples that you give, writing could take an object exactly as any so-called finite counterpart of its (and this is also how the Latin infinitive works), but writer cannot (as has been noted, writer is deverbal, not verbal).

I think the -ing form is a source of a lot of confusion in English. Setting the participle aside, the form that takes objects is considered clausal, and you can identify it also be modifiability via adverbs, and these are incompatible with determiners and pluralization:

• Writing a book
• Writing a book well/quickly etc
• * The writing a book well
• * Writings many books

These forms, proper gerunds, are analyzed as csubj when they are subjects, not nsubj, since they are not nominals. The other -ing forms are indeed deverbal nouns, and they take pluralization, determiners and adjective modifiers, but not objects or adverbs:

• The beautiful writing
• The beautiful writings
• * The beautiful writings well
• * The writing a book

The problem with labeling the latter type as verbal nouns goes back to Dan's comment about productivity above - clausal gerunds are more or less automatic, but -ing nouns are in competition with other derivation strategies (e.g. arrival is preferred to (the) arriving). Some verbs are also awkward to use with -ing derivation, or you get very different senses, for example singular "the belonging" is very odd, but a special plural sense exists for "the belongings" (which is however semantically not an automatic verbal noun describing the state of the verb "belong"). So because English has a range of competing nominalization strategies which are not totally predictable, I'm not sure it makes sense to single out nouns in -ing as verbal nouns (though it's actually not hard to find them - the tagging scheme does distinguish the different kinds of -ing).

[AngledLuffa]

Thank you to each of you for these insightful comments!

I hope I am forgiven If I self-cite my article where I also touch upon verbal nouns (in fact what sparked it was a study on Latin so-called gerund/ive).

No, that's very helpful, glad to have the citation. We can also include use that citation in any work further explores a Vnoun, such as for a specific language which might have a use for it

for the only reason that that change in denomination sparked furious and irreconcilable debates in the annotating community

Ah! It must have been a day that ended in "y"

As usual, once a form has been identified as a verbal noun it should always be annotated as VERB and not oscillate between, say, VERB and NOUN (now, incidentally, in English things look more complicated as there seem to be two formally convergent forms in -ing...).

This is actually not strictly followed in the treebanks I looked through. The Turkic ones use VERB with Vnoun, Polish uses Vnoun with NOUN, Czech uses NOUN, Bengali, Pashto, and Sinhala use NOUN, and Malayalam uses VERB. So it would seem that tagging verbal nouns as either VERB or NOUN is not standardized across treebanks.