For my constituent questions library, I want to be able to cover examples like (1):
(1) Skolko knig prochital Ivan? how.many . book.PART . read.PST . Ivan.NOM `How many books did Ivan read?' [rus]
The partitive case is interesting because with the how.many or just many word (or also this/that, actually), any noun can appear in partitive and any transitive verb can take the whole thing as its direct argument. However without the determiners, only some verbs will take the partitive argument, and only certain semantic kind of argument:
(2) Ivan vypil chaju Ivan.NOM drink.PST tea.PART `Ivan drank some tea' (3) *Ivan videl chaju Ivan.NOM see.PST tea.PART Intended: `Ivan saw some tea' (4) Ivan videl mnogo chaju Ivan.NOM see.PST many tea.PART `Ivan saw lots of tea' (5) #Ivan vypil knigi Ivan.NOM drink.PST book.PART Intended: `Ivan drank some of the book' (???) (6) *Ivan videl knigi Ivan.NOM see.PST book.PART Intended: `Ivan saw some of the book'
For now, I said that my bare-NP rule will only apply to non-partitive case things, and the only way to get a partitive NP is by the HEAD-SPEC rule. However this rules out (2).
There is a clear difference between tea and books in that tea is mass and book is countable. So I could approach this from that end, I suppose? Say that verbs like drink want mass argument, and that bare-NP applies to either to mass things OR to non-partitive things?
Does this sound like a reasonable track? The cost appears to be:
- One additional phrase-structure rule (the bare NP for mass nouns which does not restrict case)
- A new feature MASS bool which every noun and pronoun needs to account for.
But I don’t think this alone solves everything: it admits (3). The only gain compared to simply saying that all verbs take acc+part objects is ruling out (6). And while I can say that drink takes MASS + objects, it does not make sense to say that see doesn’t.