[developers] tsdb issue

R. Bergmair rb432 at cam.ac.uk
Mon Feb 16 17:50:17 CET 2009


On Mon, 16 Feb 2009, Stephan Oepen wrote:

> exactly which version of the ERG do you have, i.e. which SVN URL did
> you use?

I'm using http://svn.emmtee.net/erg/trunk

>  in case you want to experiment with the forthcoming release
> of the ERG (version 0902, aka feb-09),

Yes.  Dan and Ann tell me that it should be stable enough by now,
that it would be possible for me to treebank the data I need now,
and not have to re-treebank anything later, when the new version
is "officially" released.

> there are quite a few revisions
> to core assumptions in this release, where we are still ironing out the
> fine points of end-to-end software support.

Yes, so that's why I'd rather work with the experimental version
now than build my software (which works off of the MRSes) around
the old assumptions.

> same time, this version uses the new chart mapping functionality (see
> Adolphs et al. (2008) for background), thus requires PET binaries from
> the so-called `cm' branch.  pre-compiled binaries and [incr tsdb()] cpu
> definitions are included with the LOGON tree (see LogonTop and PetInput
> in the wiki).

Yes, arrival of the new chart mapping preprocessor in PET is
also eagerly anticipated from my side...  The particular setup
I was using for the treebanking today was using the LKB, though.
Do you think that is what might be causing my particular
"amnesia" symptom?  Are you saying that I will need to use PET
instead for the new grammar?  Will the grammar not remain
compatible with the old preprocessing architecture?  Is the
chart-mapping preprocessor (currently / going to be) implemented
in the LKB as well, or just in PET?

> we plan to email more about the exact nature of changes once this ERG
> release is complete.  if you decide to start experimentation before we
> get to that point (which should be before the end of this month :-) and
> still experience the amnesia problem with the LOGON tree, are there any
> messages printed into the *common-lisp* buffer or to your terminal?  i
> would obviously hate to see [incr tsdb()] suffer from amnesia, as it is
> barely in its teenage years now :-)!

This is what I'm currently seeing:

   [16:30:13] browse-tree(): `fracas-new' --- item # 4
   [16:30:13] browse-tree(): retrieved item # 4 (2 parses).
   [16:30:13] browse-tree(): retrieved 1 tree record.
   [16:30:13] browse-tree(): reconstructed 2 edges.
   [16:30:13] browse-tree(): retrieved 6 decisions.
   [16:30:13] set-up-compare-frame(): found 5 discriminants.

   [4] `Some Italian men are great tenors.'

I'll get an LKB source tree from the SVN now, and let you know
what happens with that.



More information about the developers mailing list