How does the disambiguator have to be adapted to the grammarchecker to be more robust?

  • we would like to use the general disambiguator as there is maintanance going on, but it does not seem adequate for sentences with e.g. case errors
  • where does it not seem adequate?
    • choosing numeral attributes: if there is a case error in the numeral, the rules still pick the attribute form instead of the correct non-attributive form (nominative/accusative)
    • postposition/adverb disambiguation: if there is a case error in the noun prior to a postposition the postposition is analyzed as an adverb
    • object analysis

Rule after Duommás modifications:

SELECT:r2599 Com IF (0 Num)(NEGATE *1 PL-NOUN BARRIER NOT-ADJ LINK 0 Pl)(NEGATE 1 Gen OR Acc LINK 0 Sem/Measr); # Cf. rules for PL-NOUN above.
	#GRAM: added: (NEGATE 1 Gen LINK 0 Sem/Measr) 
	#$ Dán jagi lea juste beannot mettara eanet muohta go diibmá
	 bohccuid , sáhttet orohatstivra ja soaitá guovlustivra eaiggáda rehkega 	alde 	mearridit háhkat dárbbašlaš geahččanveahki . 

Without Duommás change, we get the following analysis:

	"beannot" Num Sg Com Attr @>N SELECT:12217:r2599 MAP:15266:r198 &msyn-acc-to-comp ADD:4405:acc-to-comp 
;	"beannot" Num Sg Acc REMOVE:10644:r2186 
;	"beannot" Num Sg Loc Attr SELECT:12217:r2599 
;	"beannot" Num Sg Nom SELECT:12217:r2599 
;	"beannot" Num Sg Gen SELECT:12217:r2599 
;	"beannot" Num Sg Ill Attr SELECT:12217:r2599 

We need a statical disambiguator because we need to trust it in our grammarchecker product