[itsdb] MRSes in the results

Stephan Oepen oe at csli.Stanford.EDU
Wed Feb 21 06:40:01 CET 2007


hi stefan,

> I have a problem with my MRSes. I send all of them to [incr TSDB()],
> but only some are shown when I display the results in the podium. I
> guess there is a problem with the syntax of the MRS I send. Is there
> a way to let \tsdb inform me about such problems? MRS that are stored
> and displayed by the LKB MRS-viewer have some `nil's in the HCONS
> list (together with QEQ and some handles). Where could that come
> from?

apologies for a tardy reply (i am cleaning my inbox).  [incr tsdb()]
invokes whatever function is registered as the reader for the `mrs'
attribute, and whenever its return value is non-nil will display it.
the default reader function is mrs::read-mrs-from-string().  i guess
for those cases where you see no results, it returns nil.  you could
debug the reader, doing something like

  (with-input-from-string (stream "[ LTOP h0, INDEX x0, RELS < ... > ]")
    (read-mrs stream))

inserting the actual MRS string as recorded in the database.

for the specific problem of QEQs with nil arguments, these are likely
caused by the actual variables not occuring in the actual RELs.  the
code assumes that, when reading HCONS, it can only refer to variables
that have been read in at least one actual EP already (which i suspect
makes sense, as QEQs should relate to existing elements).

but the MRS manipulation code invoked by [incr tsdb()] really is part
of the MRS module of the LKB, i.e. largely written by ann.  you might
want to copy follow-up questions to the `lkb' list.

                                                          best  -  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; oepen at idi.ntnu.no ---
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++



More information about the itsdb mailing list