[developers] questions about PET/ERG system

Stephan Oepen oe at ifi.uio.no
Wed Apr 8 18:51:24 CEST 2015


>   cheap -repp -cm -default-les=all -packing \
>     -nsolutions=1 -mrs -memlimit=4096 -timeout=60 english.grm

apologies, i forgot one thing: unless working in a closed-vocabulary
domain that is fully covered by the built-in ERG lexicon and its
lightweight named entity recognition (numbers, dates, capitalized
proper names, and such; recognized by surface patterns during token
mapping), most use cases also enable unknown word instantiation based
on PoS tags, by adding ‘-tagger’ to the cheap command line.  this
option will make PET call out to a tagger, by default using the
configuration in ‘pet/common.set’:

  taggers := tnt.
  tnt-command := "${LOGONROOT}/bin/tnt".
  tnt-arguments := "-v0 -z100 ${LOGONROOT}/coli/tnt/models/wsj -".

looking at the excessive use of comments in the above file, i realize
that too might have some documentation value.

best, oe



More information about the developers mailing list