[developers] semi.vpm and trigger rules (and KNOPPIX+LKB)
sweaglesw at sweaglesw.org
Thu May 23 06:46:33 CEST 2013
Hi Emily and Sanghoun,
I verified that the problem occurs with matrix grammars both in KNOPPIX+LKB and in the main LOGON tree. Sanghoun's suggestion of commenting out the "* >> u" mapping in the VPM did not do anything for me.
However, after considerable head scratching, I came across what I believe is the real solution: the "lkb/script" file produced by the matrix customization system puts the command to load the VPM *after* the command that loads the trigger rules. These commands need to happen in the opposite order (presumably because the logic that loads transfer rules uses the VPM). I moved the VPM loading command to immediately before the trigger rule loading command, and the problem was fixed. I also see the ERG's lkb/script loads the VPM before the trigger rules.
On May 22, 2013, at 8:09 PM, Emily M. Bender wrote:
> Thanks, Sanghoun. This confirms my suspicion that the variable
> types are the problem, but I thought we just decided that
> (setf *variable-type-mapping* :semi)
> was correct. Should we back out of that, and have the semi
> do event <> e, etc?
> On Wed, May 22, 2013 at 7:03 PM, Sanghoun Song <sanghoun at uw.edu> wrote:
>> As I see, if either of them (or both) appears, trigger.mtr does not
>> work correctly.
>> (i) * >> u in semi.vpm
>> (ii) (setf *variable-type-mapping* :semi) in lkb/mrsglobals.lsp
>> For instance, if the input sentence is [犬 が 吠える], and both of them
>> above are removed, three sentences are generated (犬 が 吠える / 犬 は 吠える /
>> 犬 吠える). But, if at least one of them exists, only [犬 吠える] (without
>> semantically empty adpositions) is generated.
>> On Wed, May 22, 2013 at 6:50 PM, Sanghoun Song <sanghoun at uw.edu> wrote:
>>> Dear All,
>>> I tested this on my ubuntu machine (not KNOPPIX+LKB) with my toy
>>> Japanese grammar, and found that trigger rules do not work properly
>>> either. I also found that the following line in semi.vpm causes the
>>> * >> u
>>> If the line is removed, the generation works correctly with trigger.mtr.
>>> As a test, I added the following two lines into lkb/mrsglobals.lsp,
>>> but the trigger rules still do not work.
>>> (setf *show-lnk-p* nil)
>>> (setf *variable-type-mapping* :semi)
>>> On Wed, May 22, 2013 at 5:24 PM, Emily M. Bender <ebender at uw.edu> wrote:
>>>> Dear all,
>>>> Another discovery from working with MT in Ling 567
>>>> this week:
>>>> Once we turn on the vpm, with the following line in lkb/script:
>>>> (mt:read-vpm (lkb-pathname (parent-directory) "semi.vpm") :semi)
>>>> ... our trigger rules stop working. The grammar still loads fine, but
>>>> the trigger rules don't match and the semantically empty entries aren't
>>>> triggered. We suspect its because they reference "event", "handle", etc.
>>>> but the post-VPM MRSs have "e", "h", et al instead.
>>>> We noticed that the trigger rules in the ERG and Jacy reference
>>>> the types event, handle, etc and so suspect there is some setting
>>>> somewhere that tells the trigger rules to work on the post-VPM,
>>>> grammar-internal MRS that we are missing. Is there? If not,
>>>> any ideas on what else it might be?
>>>> p.s. I haven't had a chance to confirm yet whether this is
>>>> specific to KNOPPIX+LKB or also true in the LOGON distribution.
>>>> Emily M. Bender
>>>> Associate Professor
>>>> Department of Linguistics
>>>> Check out CLMS on facebook! http://www.facebook.com/uwclma
>>> Sanghoun Song (PhC)
>>> Dept. of Linguistics, Univ. of Washington
>> Sanghoun Song (PhC)
>> Dept. of Linguistics, Univ. of Washington
> Emily M. Bender
> Associate Professor
> Department of Linguistics
> Check out CLMS on facebook! http://www.facebook.com/uwclma
More information about the developers