mosesdecoder/contrib/m4m/modules/fastalign.m4m

# -*- makefile -*-
# M4M module for word alignment with fast_align
# see http://aclweb.org/anthology-new/N/N13/N13-1073 (paper)
# see https://github.com/clab/fast_align (github)

fstaln   ?= ${WDIR}/crp/trn/aln/fast
fstaln.in = $(addprefix ${WDIR}/crp/trn/pll/clean/, ${pllshards})
# symal 
symal_grow_diag_final_and = -a=g -d=yes -f=yes -b=yes 
symal_args = ${symal_grow_diag_final_and}

fastalign: | pll-ready
fastalign: options = -d -v -o 
fastalign: $(fstaln)/${L1}.txt.gz
fastalign: $(fstaln)/${L2}.txt.gz
fastalign: $(fstaln)/${L1}-${L2}.symal.gz

.INTERMEDIATE: $(fstaln)/tmp/${L1}-${L2}.fwd.gz
.INTERMEDIATE: $(fstaln)/tmp/${L1}-${L2}.bwd.gz
$(fstaln)/${L1}-${L2}.symal.gz: | $(fstaln)/tmp/${L1}-${L2}.fwd.gz
$(fstaln)/${L1}-${L2}.symal.gz: | $(fstaln)/tmp/${L1}-${L2}.bwd.gz
$(fstaln)/${L1}-${L2}.symal.gz: | $(fstaln)/${L1}.txt.gz
$(fstaln)/${L1}-${L2}.symal.gz: | $(fstaln)/${L2}.txt.gz
	$(lock)
	${m4mdir}/scripts/fast-align2bal.py \
		<(zcat $(fstaln)/${L1}.txt.gz) \
		<(zcat $(fstaln)/${L2}.txt.gz) \
		<(zcat $(fstaln)/tmp/${L1}-${L2}.fwd.gz) \
		<(zcat $(fstaln)/tmp/${L1}-${L2}.bwd.gz) \
		| ${symal} ${symal_args} | gzip > $@_ && mv $@_ $@
	rm -rf $(fstaln)/tmp
	$(unlock)

.INTERMEDIATE: $(fstaln)/tmp/${L1}-${L2}.txt
$(fstaln)/tmp/${L1}-${L2}.txt: | $(fstaln)/${L1}.txt.gz
$(fstaln)/tmp/${L1}-${L2}.txt: | $(fstaln)/${L2}.txt.gz
	$(lock)
	paste -d '\t' <(zcat $(fstaln)/${L1}.txt.gz) <(zcat $(fstaln)/${L2}.txt.gz) \
		| perl -pe 's/\t/ \|\|\| /' > $@_ && mv $@_ $@
	$(unlock)

$(fstaln)/tmp/${L1}-${L2}.fwd.gz: options ?= 
$(fstaln)/tmp/${L1}-${L2}.fwd.gz: | $(fast_align)
$(fstaln)/tmp/${L1}-${L2}.fwd.gz: | $(fstaln)/tmp/${L1}-${L2}.txt
$(fstaln)/tmp/${L1}-${L2}.fwd.gz: 
	$(lock)
	${fast_align} -i ${@D}/${L1}-${L2}.txt ${options} | gzip > $@_ && mv $@_ $@
	$(unlock)

$(fstaln)/tmp/${L1}-${L2}.bwd.gz: options ?= 
$(fstaln)/tmp/${L1}-${L2}.bwd.gz: | $(fast_align)
$(fstaln)/tmp/${L1}-${L2}.bwd.gz: | $(fstaln)/tmp/${L1}-${L2}.txt
$(fstaln)/tmp/${L1}-${L2}.bwd.gz: 
	$(lock)
	${fast_align} -r -i ${@D}/${L1}-${L2}.txt ${options} | gzip > $@_ && mv $@_ $@
	$(unlock)

$(fstaln)/${L2}.txt.gz: | $(addsuffix .${L2}.gz, ${fstaln.in})
	$(lock)
ifeq ($(words ${pllshards}),1)
	@cp -l $| $@ || cp $| $@
else
	@cat $| > $@_ && mv $@_ $@
endif
	$(unlock)

$(fstaln)/${L1}.txt.gz: | $(addsuffix .${L1}.gz, ${fstaln.in})
	$(lock)
ifeq ($(words ${pllshards}),1)
	@cp -l $| $@ || cp $| $@
else
	@cat $| > $@_ && mv $@_ $@
endif
	$(unlock)

# install fast-align if you don't have it
fast-align.git = https://github.com/clab/fast_align.git
${fast_align}:
	$(lock)
	git clone ${fast-align.git}
	cd fast_align && make
	mkdir -p $(dir ${fast_align})
	cp fast_align/fast_align $(dir ${fast_align})
	rm -rf fast_align
	$(unlock)