Imperative mood on auxiliaries

UniversalDependencies / UD_English-EWT

English data

Creative Commons Attribution Share Alike 4.0 International

199 stars 42 forks source link

Imperative mood on auxiliaries #326

Open nschneid opened 2 years ago

nschneid commented 2 years ago

EWT marks Mood=Imp on "do" auxiliaries when used in an imperative clause. Is this correct, and if so, what should be the features of the main verb—Mood=Imp|VerbForm=Fin or VerbForm=Inf?

EWT uses Mood=Imp|VerbForm=Fin on the main verb in addition to the AUX: http://universal.grew.fr/?custom=62a6a0f450257

With "do" plus a passive auxiliary, both are marked imperative: http://universal.grew.fr/?custom=62a6a35dc5cd7

By contrast, GUM uses Mood=Imp only on VERBs, never AUXes: http://universal.grew.fr/?custom=62a6a158e7d34 This seems like a problem at minimum because it excludes imperative uses of "be". For example: http://universal.grew.fr/?custom=62a776d81d8ad (also one token of passive "get": http://universal.grew.fr/?custom=62a6a5f10cb32)

amir-zeldes commented 2 years ago

Agreed, imperative 'be' should receive Mood=Imp. There is only one case in GUM ATM, but that's because the morph script quashes the other cases, will fix.

amir-zeldes commented 2 years ago

About the questions above:

Imperative aux do should be Imp IMO (why not?)
In general I think only the aux should have a Mood if the lexical verb is an infinitive
Similarly in the passive double aux case, only "do" should be Imp (the other aux is an infinitive, so it does not have a mood in the morphological sense).

This is not only historically and traditionally correct (we put Imp only on the verbs which would carry Imp mood in the IE sense), but also makes sure that counting Imp in the corpus results in the correct number.

nschneid commented 2 years ago

Agreed, imperative 'be' should receive Mood=Imp. There is only one case in GUM ATM, but that's because the morph script quashes the other cases, will fix.

(just fixed one of my GUM links above)

nschneid commented 2 years ago

The above policy seems reasonable if we are marking Mood=Imp on verbs. Without thinking about other languages it is hard to intuit which verb/aux should receive that designation, because in English the is no distinction between the imperative form and the plain form (i.e. VB in PTB). So one could make the argument that in English, imperative is properly a clause type, not a morphological feature (cf. UniversalDependencies/docs#877). But the "infinitive" terminology of VerbForm=Inf would be counterintuitive here because we say imperative clauses are finite, and I don't know how else to indicate the plain form of the verb.

amir-zeldes commented 2 years ago

A lot of English morphology is stricty speaking 'hypothetical' and practically speaking simply a continuation of Indo-European linguistic terminology and tradition. But there are advantages to keeping this tradition: it's probably the least confusing to non-UD linguists new to the framework, and it has the benefit of being comparable to related languages, where the IE system is more overtly in evidence. Historically speaking, only the subjunctive and imperative 'be', as wel as the inflected modals, are overt reflexes of the IE mood system, but systemically I think it makes sense to use Imp for lexical verbs where 'be' could stand if the verb were 'be' (and historically, these used to be marked in older English). Infinitives do not have mood in the IE system in any language can think of (or incidentally in its nearest neighbor, Afro-Asiatic, either), so diverging from that would probably be unexpected.