[developers] Latest ERG not generating unknown names with ACE

Michael Wayne Goodman goodmami at uw.edu
Sat Apr 15 20:26:24 CEST 2017

Thanks Stephan and Dan; replies below...

On Sat, Apr 15, 2017 at 1:17 AM, Dan Flickinger <danf at stanford.edu> wrote:

> I have not yet made any big change in the analysis of proper names, and I
> am also mystified at why the unknown name no longer gets printed in the
> generated sentence string.  What is especially puzzling is that if you
> generate with the graphical tools enabled,
> ace -g erg.dat -elp --show-gen-chart
> you'll see the unknown name cheerfully showing up both in the generator
> chart and in the parse trees for the generated sentences.  So the feature
> structures must be very nearly right, but somehow not quite right.  I'll
> keep looking through recent changes I've made to the ERG to see if I can
> find the relevant one.

Thanks Dan. It seems that it's copying the ORTH value from
generic_proper_ne in gle.tdl, but I'm not sure what would cause that. see
below for more.

>  Dan
> ------------------------------
> *From:* developers-bounces at emmtee.net <developers-bounces at emmtee.net> on
> behalf of Stephan Oepen <oe at ifi.uio.no>
> *Sent:* Saturday, April 15, 2017 12:59 AM
> *To:* Francis Bond; Michael Wayne Goodman
> *Cc:* developers at delph-in.net
> *Subject:* Re: [developers] Latest ERG not generating unknown names with
>> Unity, _generic_proper_ne_ is dropping.
>> _generic_proper_ne_ is dropping unity.
> it looks as if the right (generic) lexical entry is triggered, but its
> surface form does not reflect the CARG.  could you email the input MRS?
> what happens in the LKB?  e.g. when you do something like:
> (lkb::generate-from-mrs (mrs::read-mrs-from-file
> "~/src/logon/lingo/erg/mrs/nn.mrs"))

    TSNLP(3): (lkb::generate-from-mrs (mrs::read-mrs-from-file
    Error: Received signal number 7 (Bus error)

I'm not sure what this error means, but I don't think it's relevant to the
problem at hand. Note, the grammar *is* loaded and I can parse with
`(lkb::do-parse-tty "Dogs bark.")`.

> maybe dan has finally changed his analysis of proper names, to make them
> more similar to common nouns (no CARG)?
> —a general comment: do you guys have a specific reason to work off the ERG
> trunk?  among other things, its SEM-I (and hence generation behavior) is
> not nearly as well-defined as in the 1214 release.

I was under the impression that trunk would be the most up-to-date version
(i.e., that development is linear and 1214 is just a tagged point in time
on that development path), but I guess trunk and the 1214 branch have
diverged? Will trunk get 1214 features (like the new SEM-I) ported over?

The latest LOGON version of ERG (which is 1214, I think), does generate
(except "Unity" is not recognized as a proper name and is thus lowercased,
but that's a smaller issue):

~/logon/lingo/erg$ ace -g erg-logon.dat -e < unity.mrs
Ubuntu is dropping unity.

I attach the relevant MRSs (same sentence; one created with the ERG trunk
and the other with the 1214 version).

Michael Wayne Goodman
Ph.D. Candidate, UW Linguistics
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20170415/a0404ea0/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: unity-trunk.mrs
Type: application/octet-stream
Size: 483 bytes
Desc: not available
URL: <http://lists.delph-in.net/archives/developers/attachments/20170415/a0404ea0/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: unity-logon.mrs
Type: application/octet-stream
Size: 448 bytes
Desc: not available
URL: <http://lists.delph-in.net/archives/developers/attachments/20170415/a0404ea0/attachment-0001.obj>

More information about the developers mailing list