GrammarSoft ApS

'Chipping off'

The second, main, algorithm does the 'normal' Constraint Grammar processing, removing and selecting readings as the context becomes more and more disambiguated. The first, and largest, part of this algorithm handles cases mapped to subjects. The rules are grouped after which case they , and whether they select or remove a case.⁹ I will in the following explain each rule. The rules are usually doubled to handle both non-inverted and inverted subjects, that will be found respectively to the left and the right of their finite verb, in the same way as the rules described in last section. For each of these rule-pairs there is also an extra rule to handle subjects to the right of the finite verb, but to the left of the main verb, i.e. inversions in complex tenses like 'Derfor kan Rødhætte ikke genkende sin bedstemor' (litt.: Therefore can Little Red Riding Hood not recognise her grandmother'). Extra rules (usually one) are also added to handle up to one embedded clause (i.e. one relative clause with no embedded clause) modifying a non-inverted subject.¹⁰These two kind of rules (inversion in complex tense and relative clause) ought also be written for each of the idiosyncratic verb uses above, but the likelihood of these specific constructions occuring has been deemed to small.

• Relative clauses
• Theme
• Patient and Resultative
• Experiencer
• Objects