$Date: 2005/02/24 01:20:01 $

<hr>

From Emily Bender:
<p>

The main site for the LOGON project is 
<a href="http://www.emmtee.net">here</a>
<p>

The slides I was presenting from are available (and are
already publically linked), 
<a href="http://www.emmtee.net/tmp/tmi04.pdf">here</a>:
<p>

Most of the work on deep processing with HPSG can be
found through this portal, including links to the various
grammars: see <a href="http://www.delph-in.net">here</a>
<p>

In some cases, the most up-to-date information is in 
the wiki, which is not easy to find from the main site:
see <a href="http://wiki.delph-in.net">here</a>
<p>

I got some answers back from Stephan Oepen:
<p>
1) Their current system has about 100 tranfer rule types
and 850 instances.  That's about 1/3 more than the system
had at the time of the TMI presentation last year.
<p>
2) The results they showed comparing March and August 2004
were in fact on the dev set.  There are 104 Norwegian sentences,
but 320 English sentences, because each Norwegian sentence has
on average 3 English translations.
<p>
Lastly, The last line of the paper is apparently a traditional fairy-tale
end (like "They lived happily ever after", though I don't know what
the literal translation is).


<hr>