[developers] matrix predicate names

Emily M. Bender ebender at uw.edu
Tue Mar 1 16:45:17 CET 2016


Thanks, Francis. The ones with the old string syntax (' ) have been
updated to regular strings already (so more recently than the start
of zhong, I guess!).

As for the other ones, it is not clear to me what the desirable values
are, actually:

* named_rel follow(s|ed) the ERG, and I suspect norm_rel did too.
* the Matrix does not have a strong position on hierarchies of predicate
names, and so I think defaults to strings there.

I can see, however, that having these as strings would get in the way
for grammars that wanted types (or different strings), and also that
using "exist_q_rel" in that particular case is imposing my view of what
the quantifier predicates should look like, without really supporting it.
I note that the bare-np-phrase type doesn't seem to give a PRED value
to its quant-relation...

I'd be interested to try to harmonize these aspects of our MRS
representations,
maybe as a SIG this summer?

Emily


On Tue, Mar 1, 2016 at 1:34 AM, Francis Bond <bond at ieee.org> wrote:

> G'day,
>
> we have five instances of '_rel'  in a recent matrix (the one in zhong)
> norm_rel := predsort.
> named_rel := norm_rel.
>   [ PRED named_rel,
>   [ PRED 'implicit_coord_rel ].
>   [ PRED 'null_coord_rel ].
>      RELS <! [ PRED "exist_q_rel", ARG0 #ind, RSTR #ltop ] !>,
>
> I think one should be named_n_rel, one exist_q_rel (not a string) and
> the others should also probably be types, or if strings
> "coord_c_implict_rel", "coord_c_null_rel", ...
>
> I don't know about norm, but I suspect he is also nouny.
>
>
> --
> Francis Bond <http://www3.ntu.edu.sg/home/fcbond/>
> Division of Linguistics and Multilingual Studies
> Nanyang Technological University
>



-- 
Emily M. Bender
Professor, Department of Linguistics
Check out CLMS on facebook! http://www.facebook.com/uwclma
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20160301/bc77a279/attachment.html>


More information about the developers mailing list