[itsdb] Problem browsing Redwoods annotations

[Stephan Oepen]

hi again, tommy,

> >   - make sure you have the correct version of the ERG loaded, i.e. the
> >     one against which your Redwoods version was built, i.e. the one in
> >     the Redwoods download directory (Jun-04 for the 5th growth).
> 
> The version I had was Nov-06.  I downloaded Jun-04, but couldn't load
> it into LKB, which instead reported this error:
> 
> Error: Attempt to take the value of the unbound variable
>    '*PSQL-LEXICON-PARAMETERS*'.
> 
> >   - after invoking `Trees | Annotate', please email me all messages in
> >     the *common-lisp* buffer, plus the following two files
> >
> >     /tmp/tsdb.debug.tommy
> >     /tmp/podium.debug.tommy
> 
> I went back to the Nov-06 version of the ERG and invoked Trees |
> Annotate last night.  When I came in this morning, it had worked!  It
> was displaying the tree for "What is?"  When I clicked Next, it took
> around half an hour and then displayed "Great".  I attach the three
> files you asked for, but presumably there's nothing wrong with them
> except the speed.

in fact, using another ERG version, i believe, explains all the issues
you were experiencing.  each Redwoods profile only records the 'recipe'
used in analyzing an utterance; everytime you ask to browse or export, 
the system has to rebuild the actual analysis from applying the recipe
to the grammar itself.  those recipes (derivation trees of identifiers
of grammar rules and lexical entries) are highly version-specific, so
only the exact grammar used in building the treebank will work.  in the
log file you sent, there are many messages of the from:

  browse-tree(): failed to reconstruct item # 18 (parse # 18).

which indicate that rebuilding an analysis (using a different version
of the ERG) failed.  every now and again, one of the old recipes might
still be compatible with a newer grammar, which is why you get to see
only a handful of items from the total set.  it also explains why you
find things slow: the system is busy retrieving items, one at a time,
trying to rebuild all derivations, failing, moving to the next item.

i will have to look at the problems loading the Jun-04 ERG version in
the current LKB. it would appear we have failed to maintain backwards
compatibility here.  i will try looking for a work-around.

                                                   more soon  -  oe

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+++ Universitetet i Oslo (IFI); Boks 1080 Blindern; 0316 Oslo; (+47) 2284 0125
+++     CSLI Stanford; Ventura Hall; Stanford, CA 94305; (+1 650) 723 0515
+++       --- oe at csli.stanford.edu; oe at ifi.uio.no; stephan at oepen.net ---
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++