What is the complementizer FORM for?

Hi! I had some difficulties with the Form value drop-down menu on the complementizer page of the customization questionnaire. I wrongly assumed that this parameter referred to the embedded verb’s form. As @olzama explained to me here, this parameter refers to the form of the complementizer. For example, one can specify that the Form of that is “that”.
I would like to share my experience with this aspect of the Clausal complements page of the questionnaire in the course of the implementation of a grammar for Portuguese. This might be helpful for other new bees like myself with an LFG background.
In Portuguese, different strategies are used in clausal complementation, e.g.:

(1) A Maria lamenta que os cachorros tenham morrido.
Mary regrets that the dogs have.SBJV;PRS;3PL die.PTCP;PST
(2) A Maria lamenta os cachorros terem morrido.
Mary regrets the dogs have.INF;3PL die.PTCP;PST
‘Mary regrets that the dogs have died.’
(3) A Maria disse que os cachorros tinham morrido.
Mary said that the dogs have;IND;IPFV;3PL die.PTCP;PST
‘Mary said that the dogs had died.’

In (1) and (3), the embedded verb is finite and the complementizer is que ‘that’. While in (1) the mood of the embedded verb must be subjunctive, in (3) it must be indicative. In (2) we have an idiosyncrasy of Portuguese, the inflected infinitive, without any overt complementizer.
My grammar didn’t produce the expected results when I assigned the Form feature the values inflected_infinitive or finite. The parser didn’t crash, but the mood selection by the main verb was not enforced.
This problem was solved leaving the Form value unspecified, as shown by @olzama. I’ve tested the grammar on a large number of grammatical sentences and ungrammatical constructions with the expected results.

2 Likes

Thanks for this summary, @leoalenc !

1 Like

Hi everybody! In PorGram, the Portuguese grammar I’m developing in collaboration with @arademaker, I’ve implemented different clausal complementation stratetegies with the customization questionnaire. Using these strategies, I’ve implemented a large variety of verb types, which are characterized by the following parameters:

  1. the type of complementizer (que ‘that’, se ‘if’ or no complementizer)
  2. the mood of the embedded verb
  3. the form of the embedded verb
  4. whether a preposition heads the complementizer que and the form of this preposition

Besides, I’ve manually implemented ditransitive and control verbs subcategorizing for a clausal complement.
The grammar can parse the expected examples with all these verbs, but it also produces a lot of spurious ambiguity.

(1) a artista lamentou que os gatos tenham matado a ratazana
the artist regretted that the cat have:PRS;SBJV;3PL killed the rat
‘The artist regretted that the cat have killed the rat.’
lamentou_que_ambig

(2) o cachorro fez que o gato dormisse
the dog make:PST:IPFV;IND;3SG that the cat sleep:PST;IPFV;SBJV;3SG
‘The dog made the cat sleep’

(3) o cachorro fez com que o gato dormisse
the dog make:PST:IPFV;IND;3SG with that the cat sleep:PST;IPFV;SBJV;3SG
‘The dog made the cat sleep’

To avoid this, I suspect I must constrain the form of the complementizer. The questionnaire enables specifying a FORM feature on complementizers. An explanatory note reads:

A complementizer’s FORM is usually just its orthography, like that of an adposition. FORM is specified in Other Features.

On the “Other Features” page, however, the FORM feature is a syntactic feature of verbs. It is possible to create new features for verbs and nouns, but not for complementizers. Does anybody know how to specify the FORM of complementizers with the questionnaire?
In order to handle control verbs I’ve introduced a COMP-FORM feature to constrain the form of prepositions heading infinitival clausal complements. I could manually extend this solution to the clausal complementation strategies defined with the questionnaire, but I was very curious whether it is possible to do that solely by means of the questionnaire.

Ah, I think it just says the feature is for verbs, but I am fairly sure that if you create some FORMs, they will be available for complementizers. I will create an issue (to fixe the questionnaire wording).

On the clausal complements page:

On the Other Features page:

1 Like

Thanks, @olzama, I’ll try your suggestion.

@olzama, as you said, when one defines orthographic FORM features for complementizers on the Other Features page, they appear on the FORM Value drop-down menu list on the Clausal Complements page. I’ve introduced these features in my grammar and could eliminate the ambiguity shown above. However, the grammar now both undergenerates and overgenerates. Examples with a matrix verb like declarar ‘declare’ requiring que ‘that’ and indicative mood on the embedded verb ceased to be parsed, see (1). Instead, the ungrammatical counterparts of these examples and analogous ungrammatical examples with subjunctive mood are parsed, see (2) and (3).

(1) o cachorro declarou que o gato late
the dog declared that the cat bark:PRS:IND:3SG
(2) *o cachorro declarou que o gato lata
the dog declared that the cat bark:PRS:SBJV:3SG
(3) *o cachorro afirmou que o gato lata
the dog affirmed that the cat bark:PRS:SBJV:3SG

I’m quite puzzled by these results. I’ve inspected the grammar code generated by the Matrix and could not find the reason for this behavior:

Verbs of saying like declarar ‘declare’ and afirmar 'affirm" are assigned clausal complementation strategy nr. 8 which requires indicative mood. Strategy nr. 9 requires subjunctive mood:

ind-cl-verb-lex := fin-cl-verb-lex & clausal-nom-verb-lex & clausal-second-arg-trans-lex-item &
  [ SYNSEM [ LOCAL.CAT.VAL.COMPS < [ LOCAL [ CAT [ HEAD comp &
                                                        [ FORM que ],
                                                   WH.BOOL - ],
                                             CONT.HOOK.INDEX.SF prop ] ] >,
             NON-LOCAL.QUE.LIST < > ] ].

declar-verb-lex := ind-cl-verb-lex.

comps8-complementizer-lex-item := complementizer-lex-item & basic-non-wh-word-lex &
  [ SYNSEM.LOCAL [ CAT [ MC na-or--,
                         VAL.COMPS.FIRST.LOCAL [ CONT.HOOK.INDEX.E.MOOD indicative,
                                                 CAT [ MC -,
                                                       HEAD.FORM finite ] ],
                         HEAD.FORM que ],
                   CONT.HOOK.INDEX.SF prop ] ].

comps9-complementizer-lex-item := complementizer-lex-item & basic-non-wh-word-lex &
  [ SYNSEM.LOCAL [ CAT [ MC na-or--,
                         VAL.COMPS.FIRST.LOCAL [ CAT [ MC -,
                                                       HEAD.FORM finite ],
                                                 CONT.HOOK.INDEX.E.MOOD subjunctive ],
                         HEAD.FORM que ],
                   CONT.HOOK.INDEX.SF prop ] ].

que_4 := comps8-complementizer-lex-item &
  [ STEM < "que" > ].

que_5 := comps9-complementizer-lex-item &
  [ STEM < "que" > ].

Any hints on solving this problem?

Hi! I’ve found out the cause of the problem reported above: manually altered versions of the relevant types prevented the new definitions generated by the customization system from being used by the parser. The conclusion is that it is possible to assign orthographic form features to complementizers through the questionnaire, as explained by @olzama. Sorry for the confusion. It has been a growing challenge in the development of PorGram to reconcile the automatically generated code with handing coding designed to handle phenomena not covered by the system.

1 Like