Choices file with auxiliaries that take noun and adp subjects

I have a question about customization that is coming up in my inferred grammars. So there isn’t a real language in mind here: this has more to do with the way I’m inferring choices.

In the lexicon section of the questionnaire, verbs can select their subject based on case, not head type. But auxiliaries have to have a head type (adp or noun).

I’m not inferring the head type of an auxiliary’s subject, just the case. Therefore if the language has case marking adpositions, I define two auxiliaries for each, one that takes and adp and one that takes a noun. For example:
aux1_name=aux1
aux1_sem=add-pred
aux1_subj=np-comp-case
aux1_stem1_orth=ba
aux1_stem1_pred=_eat-or-say_v_rel
aux2_name=aux2
aux2_sem=add-pred
aux2_subj=adp
aux2_stem1_orth=ba
aux2_stem1_pred=_eat-or-say_v_rel

This choices file passes validation. However, in customization, the resulting grammar lists two head types for the subj, rather than using the disjunctive type +np:

subj-raise-aux := aux-lex & trans-first-arg-raising-lex-item &
[ SYNSEM.LOCAL [ CAT.VAL [ SPR < >,
SPEC < >,
COMPS < #comps >,
SUBJ < #subj > ],
CONT.HOOK.XARG #xarg ],
ARG-ST < #subj &
[ LOCAL [ CAT [ HEAD noun & adp &
[ CASE #case ],
VAL [ SUBJ < >,
SPR < >,
SPEC < >,
COMPS < > ] ],
CONT.HOOK.INDEX #xarg ] ],
#comps &
[ LOCAL.CAT [ VAL [ SUBJ < unexpressed &
[ LOCAL.CAT.HEAD.CASE #case ] >,
COMPS < >,
SPR < >,
SPEC < > ],
HEAD verb ] ] > ].

As a result the grammar doesn’t load.

Should the customization be able to handle auxiliaries where subjects of some cases are nouns and others are adpositions? I can’t think of an example of this, but it seems typologically plausible. If so, I should change the customization code to add HEAD +np, rather than HEAD noun & adp.

Or am I making a mistake in the choices file?

1 Like

It sounds like it might be a little of both. That is, we shouldn’t customize a grammar that won’t load, so there’s either a customization error here or a validation error, and I’m guessing the former.

But it also seems that you shouldn’t have to define two separate auxiliaries in this situation. Ideally, we’d just be able to say: its an auxiliary whose subject matches the complement’s subject in HEAD and CASE value. A language that has optional case marking adps should already allow for +np subjects…

But it also seems that you shouldn’t have to define two separate auxiliaries in this situation. Ideally, we’d just be able to say: its an auxiliary whose subject matches the complement’s subject in HEAD and CASE value. A language that has optional case marking adps should already allow for +np subjects…

How would I specify this in the questionnaire? It asks for a HEAD type for each auxiliary:
image

Or are you saying this would be ideal, but it’s not possible… so I do need to specify two separate auxiliaries?

Hmm – my guess is that this questionnaire shouldn’t be talking about HEAD type at all. Can you look at what happens in the (non-aux) verb valence frames when a language has case marking adps?

Regular verbs use +np. …so I checked what would happen if I didn’t create the duplicate entries and selected the “noun phrase bearing the case the verbal complement assigns to its subject” choice. And that puts +np on the aux.

So it seems that the functionality is good, and the questionnaire is just a little misleading (or I just misunderstood the questionnaire).

That’s good! If you have a moment, can you file a ticket to improve the questionnaire wording and/or put a note about this in the relevant MatrixDoc page?

Ticket filed!