[developers] Punctuation and "-default-les" type mapping in PET/ERG
Peter.Adolphs at dfki.de
Wed Apr 16 20:26:56 CEST 2008
R. Bergmair wrote:
> After experimenting with this some more, it now seems to me like
> PET is actually trying to unify CARGS into the FS for each word,
> *regardless* of whether the word has been instantiated from the
> posmapping as a generic type, or whether the word is actually
> coming out of the ERG lexicon.
Yes. This behaviour is triggered by the SMAF tokenizer, and finally by your
smaf.conf. It is independent of the `mrs-carg-path' mechanism that Dan referred to.
Do you need to stamp the CARG of MRSes for native lexical entries? If not, you
can get rid of the warning messages if you don't define gMap.carg and don't
refer to it in the rule for mapping pos-edges. So your example should parse
without any warnings if you use the following smaf.conf:
token. -> edgeType='tok' tokenStr=content
pos. -> edgeType='morph' fallback='' pos=content.tag
Does anyone know why the example and the built-in default smaf.conf tries to
stamp the surface form into the carg path for each item? (And what does "gMap"
stand for, anyway?)
> However, the type-II warnings are still there, and I'm
> absolutely clueless as to what's going on here.
I cannot reproduce these warnings with the current cheap from the main branch,
the current cvs erg and the built-in smaf.conf. But again, these warnings should
read as "failed to create dag for new path-value (<path> = <value>)". In your
case the path is set to `"Reptiles"', the value prints as `Reptiles' (but I
assume it's also a string, so its full name is also `"Reptiles"'). Since there
is no feature called `"Reptiles"' in the ERG, no feature structure can be
constructed with this path, and thus you get this error message and the lexical
entry will go into parsing unmodified.
It's not clear to me, though, how `"Reptiles"' can end up as a path. Maybe the
reason is again the smaf.conf you're using? Could you please provide this?
DFKI GmbH, Projektbuero Berlin, Rosenstrasse 2, D-10178 Berlin
Telefon: 030 / 24 31 02 - 183
Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
Firmensitz: Trippstadter Strasse 122, D-67663 Kaiserslautern
Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vorsitzender)
Dr. Walter Olthoff
Vorsitzender des Aufsichtsrats:
Prof. Dr. h.c. Hans A. Aukes
Amtsgericht Kaiserslautern, HRB 2313
More information about the developers