Changing case of the verb's the object

Back when we developed the clausal modifiers library, I had a couple of sentences in my Wambaya grammar that didn’t parse correctly (one grammatical example failed to parse and one ungrammatical example parsed) due to case change on the object in the subordinate clause. I want to see if this works now due to changes made by @curtosis 's valence change library or any other changes.

The language has an ergative absolutive pattern, but in subordinate clauses, the object of a transitive verb should be dative case. In the choices file, I added dative case to the object of the infinitive lexical rule like this:

verb-pc4_name=inflection
verb-pc4_obligatory=on
verb-pc4_order=suffix
verb-pc4_inputs=iverb, tverb
verb-pc4_lrt4_name=inf
verb-pc4_lrt4_feat1_name=form
verb-pc4_lrt4_feat1_value=INF
verb-pc4_lrt4_feat1_head=verb
verb-pc4_lrt4_feat2_name=case
verb-pc4_lrt4_feat2_value=dative
verb-pc4_lrt4_feat2_head=obj
verb-pc4_lrt4_lri1_inflecting=yes
verb-pc4_lrt4_lri1_orth=barda

I’m parsing with ace (unfortunately I don’t have lkb installed right now to do interactive debugging) and it looks like my tranisitve verb (‘yany’) can’t go through this lexical rule:

NOTE: lexemes do not span position 2 yanybarda'! NOTE: post reduction gap NOTE: ignoring yarru gany yanybarda manganymi’
NOTE: parsed 0 / 1 sentences, avg 125k, time 0.00704s

My suspicion is that the dative case in the rule is clashing with the absolutive case specified on the verb. Do we expect this to work now or was this sort of change left out of scope? Should my choices file look different to accomplish this case change?

It occurs to me that I could leave case underspecified on my transitive verb lexical entry and have case assigned by an obligatory class of lexical rules (where the infinitive rule licenses dative case and the various tense rules license absolutive). But then I would need different lexical rules for each class of transitive verbs (ergative-absolutive, ergative-allative, etc.), so that doesn’t seem like the right way to do this.

I think this is really a question for @curtosis — but my guess is that if the point is to change the case requirements, you’ll need to say something more about how the daughter’s ARG-ST/COMPS maps to the mother’s and this might require doing both valence reduction & valence augmentation (somehow preserving the index value).

And yeah, agreed, it wouldn’t be ideal to do different lexical rules for each class, just to get the default case assignments in.

Indeed! I had to look to be sure, but that’s more or less what I did for the passive in Japanese – a combination of valence reduction (subject demotion) and augmentation (object promotion), So I guess that’s technically more AND less…

verb-pc2_lrt2_name=passive-tr 
  verb-pc2_lrt2_feat1_name=case 
  verb-pc2_lrt2_feat1_value=dat
  verb-pc2_lrt2_feat1_head=obj
  verb-pc2_lrt2_feat2_name=case
  verb-pc2_lrt2_feat2_value=nom
  verb-pc2_lrt2_feat2_head=subj
  verb-pc2_lrt2_valchg1_operation=subj-dem
  verb-pc2_lrt2_valchg1_inputs=trans
  verb-pc2_lrt2_valchg1_argpos=post
  verb-pc2_lrt2_valchg2_operation=obj-prom
  verb-pc2_lrt2_valchg2_inputs=trans
  verb-pc2_lrt2_valchg2_argpos=post
  verb-pc2_lrt2_lri1_inflecting=yes
  verb-pc2_lrt2_lri1_orth=ware

That said, a more apt example might be the West Greenlandic [kal] antipassive that changes argument case without changing all the syntactic role: the subject changes from ergative to absolute, and the object becomes an instrumental adjunct to the now-intransitive verb. Unfortunately that was one of my held-out languages, and it did its job of identifying a gap in coverage.

@curtosis apologies for dropping this. I’m still a little unclear as to whether what I’m trying to do is possible. Basically, I need to change an absolutive object to a dative object. This sounds more like your West Greenlandic example than the Japanese example. Should I be able to accomplish what I’m trying to do with an object-removing rule followed by an object adding rule (and if so, how do I specify case?) or is that problematic because the index value won’t be preserved?

I tried removing an object and then adding and object with dative case as follows:
verb-pc4_inputs=iverb, tverb
verb-pc4_lrt4_name=inf
verb-pc4_lrt4_feat1_name=form
verb-pc4_lrt4_feat1_value=INF
verb-pc4_lrt4_feat1_head=verb
verb-pc4_lrt4_feat2_name=case
verb-pc4_lrt4_feat2_value=dative
verb-pc4_lrt4_feat2_head=newobj
verb-pc4_lrt4_valchg1_operation=obj-rem
verb-pc4_lrt4_valchg1_inputs=trans
verb-pc4_lrt4_valchg1_argpos=pre
verb-pc4_lrt4_valchg1_argtype=np
verb-pc4_lrt4_valchg2_operation=obj-add
verb-pc4_lrt4_valchg2_inputs=trans
verb-pc4_lrt4_valchg2_predname=object
verb-pc4_lrt4_valchg2_argpos=pre
verb-pc4_lrt4_valchg2_argtype=np
verb-pc4_lrt4_lri1_inflecting=yes
verb-pc4_lrt4_lri1_orth=barda

ACE throws an error when I try to compile the resulting grammar, so I don’t think this is right. I’m not sure promotion/demotion are the right thing to do here, since the subject isn’t changing and the object’s role isn’t changing either. Is there anything else I should try?

Can you send me the full choices, @kphowell ?

Hmm… I’ll have to fire up the code and look more closely, but off the top of my head I think you may need to have separate lrts for iverb and tverb. The validation code should be catching that, though, so maybe there’s a bug in that part. Unfortunately the valence change library is rigid about its inputs, mostly because I did it before @guyemerson’s diff-list witchcraft.

I split apart the pc into a separate one for iverbs (which don’t need valence change) and tverbs (which do). But the grammar still fails to load. I’m wondering if the issue is having both an object removing and adding rule in the same lrt isn’t possible. I emailed the choices file to you @curtosis and @olzama

1 Like