[matrix] Question from Matrix developers to grammar engineers: adpositions

[Francis Bond]

G'day,

> We plan to distinguish:
>
> 1) case marking adpositions with no predicate and no arguments
>
> 2) adpositions with a predicate which have ARG0, ARG1 and ARG2, where
>
> ARG0 is co-indexed with the index of the adposition,
> ARG1 with the external argument's index (typically the element the
> adp-phrase modifies)
> ARG2 the index of the adposition's complement.
>
> Do the two semantic adpositions described above cover all adpositions
> observed in the language(s) you work or have worked on?
> If you know of other semantic structures of adpositions, what are they? What
> are relevant examples?

Just a few quick examples from Jacy.

We also have particles that introduce a discourse relationship (topic
or focus) as well as case-marking.  So
Inu-ha hoeru

gets a parse:
inu(x)
hoeru(e,x)
topic(f,_,x)

THE dog barks

as well as one where it is external:
inu(x)
hoeru(e,_)
topic(f,e,x)

As for the dog, something barks  {could be the dog}

We also have things like
inu-dake hoeru

inu(x)
hoeru(e,x)
dake(f,e,_)

Only dogs bark.

I am not confident of what the best semantic representation is for this.

Not quite what you asked, but we have a predicate in the ERG which is
a supertype to compound_rel and at least some prepositions
(_of_p_rel).  This allows us to under-specify "Queensland University"
and "[the] University of Queensland" (and currently "Queensland's
University").   It would be good if the Matrix supported this kind of
variation.


Hope this helps,

-- 
Francis Bond <http://www3.ntu.edu.sg/home/fcbond/>
Division of Linguistics and Multilingual Studies
Nanyang Technological University