5 thoughts on “Frank Richter (Goethe University Frankfurt): Recursive adjectival modification in CLLRS”
Thank you for the talk!
I LOVE the (under-)specification on p.16. Do you have a constraint requiring that the type of the embedded functor be identical with that of the eventual functor or would you allow for mismatches?
For example for the non-overt arguments Laura talks about? For ex, we have:
eat_(eet),
but in a context like “Alex ate”, we would get:
[eat_(eet)]_(et)@(x_e),
where the filled in material would bridge the type mismatch?, for ex, you could have an operator INI of type (eet)(et) to get:
There is no requirement that the term embedded in the underspecified functor be of the same type as the functor. Such a type identity just happens to be a consequence of the assumed types of the non-logical constants for adjectives and adverbials in the analysis. For example, in lexical specifications a type mismatch between the term contribution in an underspecified functor and the ultimate functor could be used to express contextual requirements. (do you seen any applications for this?)
And yes, an operator INI as the one you describe would be possible. The semantic effect of it could even be left to meaning postulates.
one more question: everyone seems to have the wish to unify the combinatorics for adjectives as either all functional or all intersective. But we find both proposals:
You propose that all adjectives should be treated as functors. In David Lahm’s dissertation, he treats internal-readings of “different” as a (combinatorially) intersective adjective. I do the same for non-local “wrong”.
Is the problem that David and I don’t look at the modification of modifiers? or is the advantage of your proposal that you need less decomposition in the representation, i.e. you can do a lot in the meaning postulates?
The main difference is the level of abstraction: My version (following the idea of abstracting to the worst case) only says to which space of functions the modifiers belong (functions of a certain type), and that’s what goes into the semantic representations. You and David are much more specific about which functions these are for specific lexical items, e.g., they are intersective. That is the part I need to capture in meaning postulates.
Thank you for the talk!
I LOVE the (under-)specification on p.16. Do you have a constraint requiring that the type of the embedded functor be identical with that of the eventual functor or would you allow for mismatches?
For example for the non-overt arguments Laura talks about? For ex, we have:
eat_(eet),
but in a context like “Alex ate”, we would get:
[eat_(eet)]_(et)@(x_e),
where the filled in material would bridge the type mismatch?, for ex, you could have an operator INI of type (eet)(et) to get:
INI(eat_(eet))_(et)@(x_e)
There is no requirement that the term embedded in the underspecified functor be of the same type as the functor. Such a type identity just happens to be a consequence of the assumed types of the non-logical constants for adjectives and adverbials in the analysis. For example, in lexical specifications a type mismatch between the term contribution in an underspecified functor and the ultimate functor could be used to express contextual requirements. (do you seen any applications for this?)
And yes, an operator INI as the one you describe would be possible. The semantic effect of it could even be left to meaning postulates.
???
one more question: everyone seems to have the wish to unify the combinatorics for adjectives as either all functional or all intersective. But we find both proposals:
You propose that all adjectives should be treated as functors. In David Lahm’s dissertation, he treats internal-readings of “different” as a (combinatorially) intersective adjective. I do the same for non-local “wrong”.
Is the problem that David and I don’t look at the modification of modifiers? or is the advantage of your proposal that you need less decomposition in the representation, i.e. you can do a lot in the meaning postulates?
The main difference is the level of abstraction: My version (following the idea of abstracting to the worst case) only says to which space of functions the modifiers belong (functions of a certain type), and that’s what goes into the semantic representations. You and David are much more specific about which functions these are for specific lexical items, e.g., they are intersective. That is the part I need to capture in meaning postulates.