Append-list puzzle (wh questions, extracted determiners)

This question is mostly for @olzama but maybe also @guyemerson — I’m trying to debug a grammar from Ling 567 and having a hard time tracking down the bug. I suspect it’s because I don’t yet know how to navigate append-lists well.

The issue is that an NP built with the extracted-determiner-phrase is fitting in as the daughter of wh-ques. (There’s a separate bug in this grammar that’s making the QUE value of the NP be underspecified, but this is orthogonal.) I added SLASH.LIST < > to the non-head daughter of wh-ques, figuring this should keep extra-det-licensed NPs out of that spot, but it’s still unifying.

Do these two things look like they should unify, or is something actually wonky here?

Screen Shot 2021-02-13 at 5.41.25 PM

Relevant tdl:

wh-ques-phrase := basic-head-filler-phrase & interrogative-clause & head-final &
                         MC bool,
                         VAL #val,
                         HEAD verb ],
             NON-LOCAL.QUE.LIST < > ],
				      QUE.LIST < ref-ind > ],
                          LOCAL.CONT.HOOK.ICONS-KEY focus ],
                      LOCAL.CAT [ VAL #val &
                                      [ SUBJ < >,
                                        COMPS < > ],
                                  MC na-or-+ ] ] ].

extracted-det-phrase := basic-extracted-arg-phrase & head-compositional &
[ SYNSEM [ LOCAL #specloc & local &
[ CAT [ HEAD noun,
VAL [ SUBJ < >,
COMPS < >,
SPR < >,
SPEC < > ] ] ],
[ LIST < #local > ] > ],
[ LOCAL #local & local &
[ CAT [ HEAD det,
CONT.HOOK #hook ] ] >,
HOOK #hook ] ].

1 Like

Those AVMs look unifiable to me. The extracted-det-phrase one (the first AVM) is approximately:

SLASH = append(underspecified length list, 1-element list)

Since the arguments of the append are not yet fully fleshed out, the computation of the resulting list can’t be triggered yet, and so the result of the append is also just “list” for now – which can’t clash with the [ SLASH.LIST <> ] you added to the wh rule.

The underspecified list in the first position of the append most likely was the SLASH value of the extracted-det-phrase’s daughter. Any chance of tracking down why that was unconstrained?

1 Like

Yeah, on reflection, I think what I said was orthogonal really wasn’t: the rule last left the QUE value underspecified also left the SLASH underspecified, so what we have here is SLASH.LIST < > unifying with a SLASH that is < x > appended to underspecified.

I will investigate. I am pretty sure I cannot make the SLASH of the head daughter of the extracted-determiner phrase empty (without breaking something in my Russian grammar).

I don’t think that’s necessary — the problem with this grammar was that the constituent that served as the daughter of extra-det was underspecified for SLASH. I had thought that it would make sense for the non-head daughter of wh-ques to be SLASH < > (still do), but at any rate, making sure the possessive lex rule inherited from same-non-local-lex-rule fixed this.

Oh, yes. I will fix that (in my dev branch, for now, but can also merge into the trunk soon. Let me know if it makes sense to update the live site, too (I try to avoid it while the class is in session but can always do that if it feels like the risk of something going wrong is outweighed by the specific fixes).

We’re mostly not using the customization system at this point, so either way is fine! (Mostly: sometimes the students are still consulting it, but that’s not assigned.)