AUX in Finnish, where to draw the line

[rueter]

Hi @jpiitula, @flammie, @spyysalo, @jmnybl, @kailimp, @ftyers, @nikopartanen
UD documentation for the UPOS «AUX»
draws us to the thought of copula and TAMVE: Tense, Aspect, Mood, Voice, Evidentiality.

In a majority of the other Uralic languages, the verb of negation is also treated as an auxiliary, with the exception of Hungarian, where «nem» ‘not’ does not bear person or tense marking, but it does, actually, have a «ne» imperative ‘don't’. It would be nice to have similar strategies for Uralic languages in general, at least where their features converge.

I am making this note here, because the validator both online and local claimed that no AUX-s had been declared or documented, is there an explanation for this @dan-zeman?

Having to respecify the AUX-s, I approved the «olla» copula that was declared on the specication page but claimed to be unspecified. I subsequently specified the AUX verb of negation «ei», which is used as a lemma through out in Finnish, taking in the imperatives in «äl-» as well.

Shortly I will specify «olla» as AUX when it is used in expression of analytic tense, i.e., perfect tense [olen] mennyt ‘I [have] gone’.

Then, however, I was faced with the question of modal verbs: Are they AUX or VERB?

The UPOS documentation states that «Modal verbs may count as auxiliaries in some languages (English). In other languages their behavior is not too different from the main verbs and they are thus tagged VERB.»

Unlike the modal verbs of English, which have defective paradigms, Finnish modal verbs come with complete paradigms with the stipulation that the necessitive verbs, which take an oblique genetive agent täytyä ‘must, be required to’, pitää ‘should’, tarvita ‘need to’, and
oblique partitive agent verbs _ iljettää ‘feel awful about X-ing’, laiskottaa ‘be too lazy to’,..._
only occur in the third person singular, and thus, do not have complete paradigms.

Other modal verbs expressing possibility, such as
voida ‘be able to’,
_ saada ‘be allowed to’,_
taitaa ‘speculatively does’,
mahtaa ‘presumably does’,
saattaa ‘by chance might do’,
kyetä ‘be capable of X-ing’,
ehtiä ‘have time to’, keretä ‘have the time to’,
joutaa ‘(patient-oriented) it is time to’,
joutua ‘be forced to’,
jaksaa ‘have the tenacity to’,
raatsia ‘have the finances to’, raaksia ‘have the finances to’,
tarjeta ‘be warm/cold enough to’,
viitsiä ‘feel like X-ing’, kehdata ‘have the pluck to’,...
do not seem to differ syntactically from any other verb that takes an infinite verb argument.

There is no Finnish treebank that lists all of the modal verbs as AUX.
Sentence constituent ordering in Finnish is relatively free, and finite modal verbs seem to follow the same placement patterns as any other VERB.

Pekka [lähti] juoksemaan. (lit. Pekka [left] to run) ‘Pekka went for a run.’ NB. 3rd Inf (in Estonian supine [motion])
Pekka [rupeaa] juoksemaan. (lit. Pekka [begins] running) ‘Pekka will start running.’
Pekka [alkaa] juosta. (lit. Pekka [begins] running) ‘Pekka will start running.’ NB. 1st Inf
Pekka [voi] juosta. (lit. Pekka [can/is able to] run) ‘Pekka can run.’

The next set of word orders do not show up in official writing but definitely can show up elsewhere:
Kyllä Pekka juosta [voisi]. (lit. _Yes, Pekka run [could/would be able to] _) ‘Why yes, Pekka could run.’
Kyllä Pekka juoksemaan lähtee. (lit. _Yes, Pekka run [leave] _) ‘Why yes, Pekka will go for a run.’

Even necessitive verbs can change their position in a sentence.

Ei Pekankaan juosta [tarvitse]. lit. _Not Pekka.either run [needs to] _) ‘Not even Pekka [needs to] run.’

Hence, constituent ordering would not seem have any bearing on the matter.

I would suggest that modal verbs do not need to be labeled as AUX in UD, but I am definitely open to suggestions. Yes, this is a bit of a last-minute matter for v2.16.

[dan-zeman]

I am making this note here, because the validator both online and local claimed that no AUX-s had been declared or documented, is there an explanation for this @dan-zeman?

Yes, there is. There were 11 Finnish auxiliaries registered in the system until yesterday: aikoa, ei, joutua, mahtaa, olla, pitää, saattaa, taitaa, tarvita, täytyä, voida. But these had never been entered through the standard channel, i.e., the AUX registration form. Instead, they were copied as a simple list from the source code of the validator, where the auxiliaries used to be listed in the past. So they were tolerated by the validator but at the same time they were flagged as "not properly documented". Such auxiliaries only stay in the system until somebody has time to make it right. As soon as someone starts editing them, the system removes the undocumented ones and the contributor can re-introduce them in the standard way. That happened yesterday at 21:29 when you, @rueter, edited and saved olla.

[jmnybl]

The Finnish treebanks currently follow the list of 11 Finnish auxiliaries mentioned by @dan-zeman, and I would prefer not to change this, as doing so would require substantial modifications to the data without clear benefits. The only difference appears to be aikoa (“be going to”) which is annotated as AUX in TDT, OOD and PUD, but as VERB in FTB.

When that list was compiled (I believe there was an email discussion about this), we also compared it with the Estonian list, and they looked quite similar.

[rueter]

Thanks for your feedback, @dan-zeman and @jmnybl!
To make a quick patch for what happened when I opened and without alteration saved the copula «olla», I will enter the list of regular verbs that are annotated as AUX in the Finnish treebanks.
In the future, and even now, I am interested in finding the argumentation and logic behind the list of so called AUX verbs. So, if someone can find the email chain of discussion, it would be enlightening.
The list appears to contain three necessitative verbs, the ones with oblique (genitive) agents {pitää [VERB], tarvita [VERB], täytyä [VERB]}, of which the first two have AUX (necessitive: ‘should’, ‘need to’) readings, and other (‘hold [smth], keep [smth]; like [smth]’, ‘need [smth]’).
Other Uralic and Turkic languages have also listed necessitative verbs as AUX.

There are three epistemic verbs, i.e., verbs that indicate we are deducing something:
mahtaa: Mitä Liisu [mahtaa] tehdä sille? ‘What [could/might (non-past)] Liisu be doing for/about it?’
saattaa: Liisu [saattaa] ostaa jäätelön. ‘Liisu [might] buy an ice cream. ’
taitaa: Liisu [taitaa] ostaa jäätelön. ‘Liisu [is probably going to] buy an ice cream.’

The verb «voida» ‘can, have the possibility to’. It is this blank possibility which seems to leave the door open to all kinds of other modality, where possibility is specified according to kyetä, ehtiä, keretä, raatsia, raaksia, tarjeta, jaksaa...

The verb «aikoa» ‘intend to’, which indicates some kind of volition, can be contrasted with the verb «joutua» ‘be forced into’ , e.g., joutua naimisiin ‘have to get married’. The interpretation of «aikoa» ‘intend to’ as an AUX would require that its synonym «meinata» should also be AUX.

These words were never documented as AUX, so the fact that they are now being documented IN this use is not indication of their being excepted as AUX, but rather a simple quick fix.

Of the Finnic languages, Livvi, Karelian and Estonian have not documented any AUX-s other than the copula.

[rueter]

Hi @kailimp and @dan-zeman
It is interesting to see that the Estonian treebanks feature what looks to me like evidentiality verbs; something that the Finnish treebanks might not be doing. There are verbs referencing modes of acquiring information «näima», «paistma» ‘?appear to ’ and «tunduma» ‘feel like/seem like’.
Would their analogical equivalents in Finnish «näkyä», «näyttää» , «tuntua» <feel, seem>, «kuulua» , «maistua» merit treatment as AUX? In addition to expressing channel of perception, all of these verbs share the same government, i.e., the complement verb in -vAn and -neen . This is something like what we have in the Estonian -vat quotative.

I'm still searching for argumentation for how an AUX vs VERB split might be defined, @ftyers can you give some input on this?

[rueter]

Hi, @dan-zeman!
Thanks for the tags. I have now specified the necessitative verbs «pitää», «tarvita» and «täytyä», and I can put «voida» ‘be able to/ can’ under potential, BUT I am kind of stumped on what to do with the epistemic modals {mahtaa, saattaa, taitaa}.

Can you and @jmnybl make suggestions as to how to classify them and the verbs «aikoa» ‘intend to’, «joutua» ‘be forced to’

[dan-zeman]

what to do with the epistemic modals {mahtaa, saattaa, taitaa}

If these have to be AUX, I would put them under potential as well.

[rueter]

@flammie, @jpiitula and @ftyers I can imagine that when we finally sit down to define what should be called AUX in Finnish stacking might come in handy.

For example, the sentence «ei ole tarvinnut voida mennä»
(1) aux(mennä, voida)
(2) aux(mennä, tarvinnut)
(3) aux(mennä, ole)
(4) aux(mennä, ei)
provides us with a morphological side: (4) is associated with the form of (3), (3) is associated with the form of (2).

Necessitative verbs do not seem to be compatible with epistemic verbs.