running post

[Brian MacWhinney]

On 1/20/04 2:27 PM, "Deborah A. Hwa-Froelich, PhD" <hwafroda at slu.edu> wrote:

> I was wondering whether the post function has changed since June, 2003.
>

Yes, Christophe has fixed many things.  The biggest change was the
elimination of the "gaping loophole" that allowed every word with a plus
sign to be interpreted as a noun+noun compound.  Now all compounds must be
explicitly listed.  I have done a lot of work in this area, which I will
soon write up for an article.

> I ran MOR and then tried to run post but the files seem to be exactly
> the same-do I still need to run the post command to disambiguate the MOR
> line? If so, what is the command line for post?
>
> I used: posttrain B1 +0newerrors.cut filename.cex
> and I tried: posttrain B1 +0newerrors.cut filename.cha.
> and we tried: post filename.cha  None of these commands ever worked with
> an error message "No Brill's rules"
>

You want a variant of the last command.  But my guess is that you didn't put
your eng.db file into the MOR grammar folder along with eng.ar and such.


>
> Deb Hwa-Froelich
>

Deb,
  I built a new eng.db file in early December which I have been using.  It
is now included with the MOR grammar that you can download and that you set
your MOR LIB to that folder.  You need to make sure that you are running
1.  current CLAN
2.  a current ENGLISH grammar with engtags.cut
3.  current eng.db

In other words, just get everything off the web.  In that case, you can skip
POSTTRAIN.

You just run

1. mor +xl *.cha  (to make sure all words are recognized)
2. mor +1 *.cha  (to enter %mor in all files)
3. post +tengtags.cut +1 *.cha  (to disambiguate)

I have been using this combination hundreds of times now on the complete
English normal database for the last two months.  It works beautifully in
the sense that all words get disambiguated, everything passes CHECK, and the
results can also go straight into XML (which you don't need right now).

--Brian