add script for testing on Jenkins

.gitignore

@@ -36,3 +36,5 @@
 
 build
 scripts/docker
+amunmt-distribution.tar.gz
+tests/wmt16/*/*

scripts/jenkins/build.sh

@@ -8,3 +8,6 @@ mkdir build
 cd build
 cmake ..
 make
+
+cd ..
+tar zvcf amunmt-distribution.tar.gz build/bin/* scripts/download_models.py tests/wmt16/Makefile tests/wmt16/extract_segs.py

tests/wmt16/Makefile

@@ -0,0 +1,59 @@
+DOMAIN=news
+SRC=de
+TRG=en
+
+MOSES_SCRIPTS=/opt/moses/moses-scripts
+AMUNMT_BIN=../../build/bin
+
+DIRECTION=$(SRC)-$(TRG)
+DIR=$(DOMAIN)-$(DIRECTION)
+
+NUMBER_OF_LINES=99999
+
+.SECONDARY:
+
+$(DIR)/bleu.points: $(DIR)/bleu.score
+	perl -ne '/BLEU = (\d+\.\d+)/; print "BLEU\n$$1\n"' < $< > $@
+
+$(DIR)/bleu.score: $(DIR)/test-$(SRC)$(TRG)-ref.$(TRG).tok $(DIR)/test-$(SRC)$(TRG)-out.$(TRG).tok
+	$(MOSES_SCRIPTS)/generic/multi-bleu.perl $< < $(DIR)/test-$(SRC)$(TRG)-out.$(TRG).tok > $@
+
+%.$(TRG).tok: %.$(TRG).txt
+	$(MOSES_SCRIPTS)/tokenizer/tokenizer.perl -l $(TRG) -a < $< > $@
+
+%-out.$(TRG).txt: %-out.$(TRG).bpe
+	sed 's/\@\@ //g' $< | \
+	$(MOSES_SCRIPTS)/recaser/detruecase.perl | \
+	$(MOSES_SCRIPTS)/tokenizer/detokenizer.perl -l $(TRG) > $@
+
+%-out.$(TRG).bpe: %-src.$(SRC).bpe $(DIR)/config.yml
+	head -n $(NUMBER_OF_LINES) < $< | $(AMUNMT_BIN)/amun -c $(DIR)/config.yml > $@
+
+%.$(SRC).bpe: %.$(SRC).pre $(DIR)/$(SRC)$(TRG).bpe
+	$(AMUNMT_BIN)/bpe $(DIR)/$(SRC)$(TRG).bpe < $< > $@
+
+%.$(SRC).pre: %.$(SRC).txt $(DIR)/truecase-model.$(SRC)
+	$(MOSES_SCRIPTS)/tokenizer/normalize-punctuation.perl -l $(SRC) < $< | \
+	$(MOSES_SCRIPTS)/tokenizer/tokenizer.perl -l $(SRC) -a | \
+	$(MOSES_SCRIPTS)/recaser/truecase.perl -model $(DIR)/truecase-model.$(SRC) > $@
+
+
+$(DIR)/test-$(SRC)$(TRG)-ref.$(TRG).txt: $(DIR)/test-$(SRC)$(TRG)-ref.$(TRG).sgm
+	./extract_segs.py < $< > $@
+
+$(DIR)/test-$(SRC)$(TRG)-src.$(SRC).txt: $(DIR)/test-$(SRC)$(TRG)-src.$(SRC).sgm
+	./extract_segs.py < $< > $@
+
+$(DIR)/test-$(SRC)$(TRG)-ref.$(TRG).sgm: $(DIR)/test.tgz
+	tar --to-stdout -zxvf $< test/$(DOMAIN)test2016-$(SRC)$(TRG)-ref.$(TRG).sgm > $@
+
+$(DIR)/test-$(SRC)$(TRG)-src.$(SRC).sgm: $(DIR)/test.tgz
+	tar --to-stdout -zxvf $< test/$(DOMAIN)test2016-$(SRC)$(TRG)-src.$(SRC).sgm > $@
+
+$(DIR)/test.tgz:
+	mkdir -p $(DIR)
+	wget 'http://data.statmt.org/wmt16/translation-task/test.tgz' -O $@
+
+$(DIR)/model.npz $(DIR)/vocab.$(SRC).json $(DIR)/vocab.$(TRG).json $(DIR)/$(SRC)$(TRG).bpe $(DIR)/truecase-model.$(SRC) $(DIR)/config.yml: ../../scripts/download_models.py
+	mkdir -p $(DIR)
+	$< -w $(DIR) -m $(DIRECTION)

tests/wmt16/extract_segs.py

@@ -0,0 +1,10 @@
+#!/usr/bin/env python2
+# -*- coding: utf-8 -*-
+
+import re
+import sys
+
+for line in sys.stdin:
+    m = re.search(ur'<seg id="\d+">(.*)</seg>', line)
+    if m:
+        print m.group(1)