#!/usr/bin/env csh 


# Usage:
# cat $MD/DATA/SAMPLE/small.tok \
# | $MD/tsgpipe > $MD/DATA/SAMPLE/small.tsg
#
# The input to this pipe are the files in DATA/TOK.
# The output files are in DATA/TSG/*.tsg.
# N.B. DATA/TSG also contains *.gr files which are the output of
# running the TSG on DATA/TSG/*.tsg files.
#
# Difference between this and casspipe:
#
# (1) in casspipe parentheticals are marked up before tagging
# but here they are marked up afterwards - it's not clear which would
# give better tagging results. In both the parentheticals are removed 
# by the final call to xmlperl  
# (2) lemmatisation done here but not in casspipe, because tsg needs lemmas
# (3) word internal white space changed to - or _ before tagging here but
# done after tagging in casspipe. (This might have given bad results for 
# casspipe except that all words with ws have a K attribute which overrides 
# the tagger's decision).
# (4) Use of K attribute to overide tagger here is done in final xmlperl
# call, instead of during tagging.


$MD/bin/fsgmatch -q ".*/SENTENCE" $MD/GRAM/smallnumbers.gr \
| $MD/bin/fsgmatch -q ".*/SENTENCE" $MD/GRAM/hyph.gr \
| $MD/bin/sgdelmarkup -q ".*/W/W" \
| $MD/bin/sgdelmarkup -q ".*/W/W" \
| $MD/bin/xmlperl $MD/RES/clawstagger.rule \
| $MD/bin/xmlperl $MD/RES/lemma.rule \
| $MD/bin/fsgmatch -q ".*/SENTENCE" $MD/GRAM/paren.gr \
| $MD/bin/xmlperl $MD/RES/tsg.rule \
| $MD/RES/tsgstops.perl