On Thu, 8 May 2008, Ben Haskell wrote:
That's my patch that I published at
http://wordnet.princeton.edu/~ben/WN3.loop.patch
Here's a corresponding patch that I just created for the lexicographer files:
http://wordnet.princeton.edu/~ben/WN3.loop-lex.patch
Ahh, that's great. So there was just a typo in Francis' patch (supress ->
suppress).
Your patch works in fact and I'll apply it to the Debian WordNet package.
Once I found some spare minutes to patch the man pages I'll upload a fixed
WordNet package.
It does the minimum necessary to produce the same results as the other patch.
It shouldn't change the output files' offsets. Though I'm not sure I'd worry
about this on the Debian version, since you have other patches that do. I
forget whether I ever circled back to you (Andreas) on this before I left
WordNet, but is there some kind of disclaimer in the Debian version that
points out this problem?
No - and I admit that I do not really understand the consequences of the
problem. :-(
The short version: the file offsets, which are the easiest-to-use ID's in
WordNet, change anytime there's a change in the lexicographer files. The
synset keys, which are harder to compute, don't change nearly as haphazardly.
Could you elaborate on how to recalculate the synset keys? You certainly
will have any code that does this. Or would it rather be a bad idea to work
with different synset keys than your upstream database? I admit I have not
the slightest idea of the use of the synsets. I as a poor physicist use
WordNet as a really great monolingual dictionary which is very complete and
easy to use. I know that this is only a part of the intention of WordNet but
I fail to understand all the other meanings because of pure lack of
understanding
the basics of your science.
But, since they're harder to compute, no one uses them. So, anyone who wants
to use WordNet for research purposes will undoubtedly come across groups that
use the file-offset style of ID's, which, since you incorporate
Debian-specific patches, will be wrong for the Debian package of WordNet.
Thanks for this warning. I think I should at least incorporate this and if
I understood you correctly than does this concern the binary package
wordnet-sense-index exclusively which contains the index.sense file. It
was left out in former versions of the Debian package but was added on user
request as a low priority package. But this warning should be incorporated
to make users aware. Can you please confirm that index.sense is affected
but nothing else?
Do you have an other suggestion to be able to build the WordNet database
from the ASCII source using grind which would not have any negative side effect?
[will respond to the rest off the debian-bugs thread]
Ahh - I read this first because it came to my private folder ... ;-)
Many thanks for your support
Andreas.
--
http://fam-tille.de
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
