Makefile.EvalUD

ALUD=~/bin/alud
ALUD_OPTIONS=-t -d -x
EVAL=$(ALPINO_HOME)/UD2.0/conll17_ud_eval.py -v
GOODKEYS=python3 $(ALPINO_HOME)/EvalUD/goodkeys.py

compare: $(SUITES:%=%.compare)

parse: $(SUITES:%=%.log)

%.log:
	$(MAKE) annotate suite=$*

annotate:
	mkdir -p $(suite)
	$(ALPINO_HOME)/src/Alpino -notk cmdint=off\
          -veryfast\
          user_max=600000 debug=1\
          xml_format_frame=on\
          xml_format_failed_parse=on\
          display_main_parts=off\
          display_quality=off\
          -flag suite $(SUITESDIR)/$(suite)\
          -flag treebank $(suite)\
          demo=off\
          end_hook=xml $(extra)\
          $($(suite)_test1_options)\
          batch_command="parser_comparisons" 2> $(suite).log


%.conllu: %/*.xml $(ALUD)
	ls -v $*/*.xml |  $(ALUD) $(ALUD_OPTIONS) > $*.conllu 2> $*.conllu.err

%.conllu.cln: %.conllu $(CONLLU)/%.conllu
	$(GOODKEYS) $(CONLLU)/$*.conllu $*.conllu > $*.conllu.cln

%.compare: $(CONLLU)/%.conllu %.conllu.cln
	$(EVAL) $(CONLLU)/$*.conllu $*.conllu.cln > $*.compare

analyse:
	@echo "mean per corpus results; does not take into account size of corpora"
	@grep UPOS *.compare | awk '{ N=N+1; T = T + $$9 } END { print "UPOS\t" T/N }'
	@grep XPOS *.compare | awk '{ N=N+1; T = T + $$9 } END { print "XPOS\t" T/N }'
	@grep Feats *.compare | awk '{ N=N+1; T = T + $$9 } END { print "Feats\t" T/N }'
	@grep AllTags *.compare | awk '{ N=N+1; T = T + $$9 } END { print "AllTags\t" T/N }'
	@grep Lemmas *.compare | awk '{ N=N+1; T = T + $$9 } END { print "Lemmas\t" T/N }'
	@grep UAS *.compare | awk '{ N=N+1; T = T + $$9 } END { print "UAS\t" T/N }'
	@grep LAS *.compare | awk '{ N=N+1; T = T + $$9 } END { print "LAS\t" T/N }'

clean:
	rm -f  $(SUITES:%=%.compare) $(SUITES:%=%.log) $(SUITES:%=%.conllu.cln)
	rm -rf $(SUITES)