[developers] Interactive MT not working on KNOPPIX+LKB

Stephan Oepen oe at ifi.uio.no
Wed May 22 09:13:30 CEST 2013


colleagues,

i believe this is another ripple effect of ‘bogus’ LNK values
sneaking into MRSs; see:

  http://lists.delph-in.net/archives/developers/2012/001696.html

the problem was addressed in the LOGON trunk last fall:

------------------------------------------------------------------------
r11685 | oe | 2012-08-12 10:32:37 +0200 (Sun, 12 Aug 2012) | 1 line

make MRS output and input more robust (in terms of bogus LNK values)
------------------------------------------------------------------------

but it appears i have yet to promote that change into the LKB
trunk, i.e. even latest LinGO builds may still have the issue.  i
will aim to synchronize (what should only be small differences
between) the two trees before the summit.  in the meantime, i
wholeheartedly embrace the work-around found by woodley.

best, oe


On Wednesday, May 22, 2013, Emily M. Bender wrote:

> Dear Woodley,
>
> Right --- and you even pointed this out in class on Monday.
> Yes, that fixes the problem.  The dogs are now sleeping in Italian
> as well (cani dorm-ono!).
>
> I'd be interested in any insight on what changed --- I guess
> my eng and ita grammars (as well as the current customization
> system) are out of sync with some assumptions in the lkb,
> but whatever changed in the distro that we have on KNOPPIX+LKB
> isn't in the LOGON version that Ned's using?
>
> Thanks again,
> Grazie,
> Emily
>
> On Tue, May 21, 2013 at 10:07 PM, Woodley Packard <sweaglesw at sweaglesw.org
> > wrote:
>
>> Dear Emily,
>>
>> I believe the problem is the empty characterization <>'s after the EP
>> names in the MRS output from the source grammar.  Have you tried adding the
>> following to lkb/mrsglobals.lsp on the source side?
>>
>> (setf *show-lnk-p* nil)
>>
>> This worked for me, on KNOPPIX+LKB.
>> -Woodley
>>
>> On May 21, 2013, at 9:46 PM, Emily M. Bender wrote:
>>
>> Dear all,
>>
>> We've gotten to the MT portion of Ling 567 at UW, and discovered
>> that the MT set up on the KNOPPIX+LKB system isn't working.  Specifically:
>>
>> I have the students start with and eng<>ita pair that I know works.
>> (And Ned at least has it working on LOGON trunk currently.)  They
>> set up two emacsen, load the source and target grammars, and start
>> the generator server on the target side.  The values for *translate-grid*
>> are set appropriately.
>>
>> When we parse a sentence in the source grammar and select "rephrase",
>> the target side emacs reports "read 0 MRSs are generator input".  I've
>> seen this work with exactly this pair of grammars in past years (and
>> Ned is seeing it work this year), so I suspect the problem has something
>> to with the file encoding of the .translate.<user>.<iso> file.  I've
>> captured one of those by turning off the generator server, and it is
>> attached.
>> (I had to rename it to be able to upload it into gmail for some reason;
>> the actual file name was as expected.)
>>
>> Another symptom is that Generate | Display input MRS does nothing,
>> which suggests that the MRS isn't properly imported --- the grammar
>> doesn't
>> get a chance to try generating from it.  For completeness, I put the same
>> grammar (eng) on both sides (and updated *translate-grid* appropriately).
>> Same behavior.
>>
>> Here is a little bit more info:
>>
>> LKB(5): (trace mrs::read-mrss-from-file)
>> (MRS::READ-MRSS-FROM-FILE)
>> LKB(6):
>>  0[5]: (MRS::READ-MRSS-FROM-FILE
>> #P"/home/ubuntu/tmp/.transfer.ubuntu.ita")
>>  0[5]: returned NIL #<SIMPLE-ERROR @ #x4bea3d6a>
>>
>> [21:32:27] translate(): read 0 MRSs as generator input.
>> LKB(6): (trace mrs::read-mrs-stream)
>> (MRS::READ-MRS-STREAM)
>> LKB(7):
>>  0[5]: (MRS::READ-MRSS-FROM-FILE
>> #P"/home/ubuntu/tmp/.transfer.ubuntu.ita")
>>    1[5]: (MRS::READ-MRS-STREAM #<EXCL:FILE-SIMPLE-STREAM
>>
>> #P"/home/ubuntu/tmp/.transfer.ubuntu.ita" for input pos 0
>>                                  @ #x4beb10a2>)
>>    1[5]: returned-by-throwing to tag IGNORE-ERRORS-1: NIL
>>            #<SIMPLE-ERROR @ #x4beb3d12>
>>  0[5]: returned NIL #<SIMPLE-ERROR @ #x4beb3d12>
>>
>> [21:33:41] translate(): read 0 MRSs as generator input.
>> LKB(7):
>>
>> Any ideas on what might be going on here, and what kind of work-around I
>> can offer the students?
>>
>> Many thanks,
>> Emily
>>
>> --
>> Emily M. Bender
>> Associate Professor
>> Department of Linguistics
>> Check out CLMS on facebook! http://www.facebook.com/uwclma
>> <dot.transfer.ubuntu.ita>
>>
>>
>>
>
>
> --
> Emily M. Bender
> Associate 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/20130522/997df6e3/attachment-0001.html>


More information about the developers mailing list