[developers] coreference problem fixed

[Emily M. Bender]

Dear Bernd and Peter,

I have had a chance to test out the new pet and the problem does
indeed seem to be fixed.  Many thanks!  This will speed up grammar
development considerably :)

I am left with a few diffs between pet and lkb profiles at this point,
but on spot checking, it appears that those that don't relate to differing
time-out conditions all have to do with a difference in the way morphology
is handled: When there is ambiguity in the orthographic subrules,
the LKB finds multiple derivations that all involve exactly the same
lexical rules (and lexical entries and phrase structure rules).  In this
case, I'd rather not see pet adapt to what the LKB is doing.

Thanks for the warning on the other coref issue.  As long as
that doesn't also apply to cases like the following, it shouldn't
be too hard to avoid:

type1 := supertype &
 [F1 #co1 & [F2 val],
  F3 #co1 ].

type2 := type1 &
 [F1 #co2,
  F4 #co2 ].

Thanks again,
Emily

On Tue, Aug 12, 2008 at 11:00 AM, Bernd Kiefer <kiefer at dfki.de> wrote:
> Hi Emily,
>
> i think i fixed the problem that leads to different readings for LKB and
> PET with the grammar you gave Peter. He tracked the problem down and i
> could verify with a small example grammar that it now seems to work.
>
> On my way to solution, i unfortunately found another bug concerning
> coreferences in type definitions which obviously nobody found yet.
>
> TDL allows definitions of the following form:
>
> foo := [ f0 #co1 & #co2 & [ a 1 ],
>        f1 #co1,
>        f2 #co2 ].
>
> but this results in the following structure after flop'ing it:
>
> [ f0 #c1 [ a 1 ], f1 #c1, f2 *top* ]
>
> Obviously, the second coref is not treated properly. I think this error is
> hard to track down if used in a grammar, thus i wanted to warn you,
> just in case.
>
> Best,
>        Bernd
>
>