[developers] PET development discussion points

Bernd Kiefer kiefer at dfki.de
Thu Mar 20 15:51:17 CET 2008

Hi all,

i had some discussions with Peter Adolphs last week and we agreed that 
input from a greater audience is needed. There are mainly two points we 
weren't sure about:

1. the resource limits in pet, esp. in the case were unpacking is used.

we had the feeling that it is not very reasonable to spend all resources 
on "normal" parsing and do no unpacking at all if a limit is hit. 
Especially to me, it seems that it is more sensible to always leave some 
amount of the resources to the unpacking phase. Only unpacking is doing 
some kind of best first processing, which is important in real-world 
scenarios that use resource limits. My proposal for a uniform treatment 
of the different limits would be that one can specify a ratio (e.g. 20%) 
that is reserved for unpacking. I'd rather not duplicate all limit 
settings for the unpacking phase for reasons of simplicity, also for the 

And now for something completely different:

2. overwriting the stem in generic entries
This is of much interest in the new version of the preprocessor, but 
maybe should be also changed in the "traditional" version, because 
nowadays morphological processing is done on words whose lexicon entries
are generics. Is there anybody who has objections against this? 
Otherwise, we'll open a new ticket for that.

Thanks in advance for your contributions,


Bernd Kiefer       DFKI GmbH, Stuhlsatzenhausweg, D-66123 Saarbruecken
kiefer at dfki.de     +49-681/302-5301 (phone)     +49-681/3025338  (fax)
Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
Trippstadter Strasse 122, D-67663 Kaiserslautern, Germany
Geschaeftsfuehrung: Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vor-
                     sitzender), Dr. Walter Olthoff
Vorsitzender des Aufsichtsrats: Prof. Dr. h.c. Hans A. Aukes
Amtsgericht Kaiserslautern, HRB 2313

More information about the developers mailing list