2019-02-04 15:27:56 +01:00
|
|
|
SRC_LANG=pl
|
|
|
|
|
TRG_LANG=en
|
|
|
|
|
CORPUS_NAME=opensubtitles_sample
|
|
|
|
|
SEPARATOR=@\#@
|
|
|
|
|
|
|
|
|
|
DICTIONARY_WEIGHT=3
|
|
|
|
|
|
2019-02-06 15:34:08 +01:00
|
|
|
all: corpora/$(CORPUS_NAME)/falign_corpus.txt
|
2019-02-04 15:27:56 +01:00
|
|
|
|
|
|
|
|
|
2019-02-06 15:34:08 +01:00
|
|
|
clean:
|
2019-02-04 15:27:56 +01:00
|
|
|
rm -f corpora/$(CORPUS_NAME)/*.lem
|
|
|
|
|
rm -f corpora/$(CORPUS_NAME)/*.dict
|
|
|
|
|
rm -f corpora/$(CORPUS_NAME)/src_clean.txt
|
|
|
|
|
rm -f corpora/$(CORPUS_NAME)/trg_clean.txt
|
2019-02-06 15:34:08 +01:00
|
|
|
rm -f corpora/$(CORPUS_NAME)/ids_clean.txt
|
|
|
|
|
rm -f corpora/$(CORPUS_NAME)/falign_corpus.txt
|
2019-02-04 15:27:56 +01:00
|
|
|
|
|
|
|
|
corpora/$(CORPUS_NAME)/src.dict:
|
|
|
|
|
./collect_dict.py $(SRC_LANG) $(TRG_LANG) $(DICTIONARY_WEIGHT) > $@
|
|
|
|
|
|
|
|
|
|
corpora/$(CORPUS_NAME)/trg.dict:
|
|
|
|
|
./collect_dict.py $(TRG_LANG) $(SRC_LANG) $(DICTIONARY_WEIGHT) > $@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
corpora/$(CORPUS_NAME)/src.lem: corpora/$(CORPUS_NAME)/src.txt
|
|
|
|
|
/usr/local/bin/concordia-sentence-tokenizer -c ../concordia.cfg < $< | ./sentence_lemmatizer.py $(SRC_LANG) > $@
|
|
|
|
|
|
|
|
|
|
corpora/$(CORPUS_NAME)/trg.lem: corpora/$(CORPUS_NAME)/trg.txt
|
|
|
|
|
/usr/local/bin/concordia-sentence-tokenizer -c ../concordia.cfg < $< | ./sentence_lemmatizer.py $(TRG_LANG) > $@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-02-06 15:34:08 +01:00
|
|
|
corpora/$(CORPUS_NAME)/src_clean.txt corpora/$(CORPUS_NAME)/trg_clean.txt corpora/$(CORPUS_NAME)/ids_clean.txt corpora/$(CORPUS_NAME)/falign_corpus.txt: corpora/$(CORPUS_NAME)/src.txt corpora/$(CORPUS_NAME)/trg.txt corpora/$(CORPUS_NAME)/ids.txt corpora/$(CORPUS_NAME)/src.lem corpora/$(CORPUS_NAME)/trg.lem
|
|
|
|
|
./clean_corpus.py corpora/$(CORPUS_NAME)/src.txt corpora/$(CORPUS_NAME)/trg.txt corpora/$(CORPUS_NAME)/ids.txt corpora/$(CORPUS_NAME)/src.lem corpora/$(CORPUS_NAME)/trg.lem corpora/$(CORPUS_NAME)/src_clean.txt corpora/$(CORPUS_NAME)/trg_clean.txt corpora/$(CORPUS_NAME)/ids_clean.txt corpora/$(CORPUS_NAME)/falign_corpus.txt
|
2019-02-04 15:27:56 +01:00
|
|
|
|
|
|
|
|
corpora/$(CORPUS_NAME)/pasted.txt: corpora/$(CORPUS_NAME)/src_clean.txt corpora/$(CORPUS_NAME)/trg_clean.txt
|
|
|
|
|
./paste.py corpora/$(CORPUS_NAME)/src_clean.txt corpora/$(CORPUS_NAME)/trg_clean.txt $(SEPARATOR)> $@
|
|
|
|
|
|
|
|
|
|
corpora/$(CORPUS_NAME)/pasted_deduplicated.txt: corpora/$(CORPUS_NAME)/pasted.txt
|
|
|
|
|
sort -k 1.13 $< | uniq -s 12 | sort > $@
|
|
|
|
|
|
|
|
|
|
corpora/$(CORPUS_NAME)/aligned_final.txt corpora/$(CORPUS_NAME)/src_final.txt corpora/$(CORPUS_NAME)/trg_final.txt: corpora/$(CORPUS_NAME)/pasted_deduplicated.txt corpora/$(CORPUS_NAME)/aligned.txt
|
|
|
|
|
./extract.py $< corpora/$(CORPUS_NAME)/aligned.txt corpora/$(CORPUS_NAME)/aligned_final.txt corpora/$(CORPUS_NAME)/src_final.txt corpora/$(CORPUS_NAME)/trg_final.txt $(SEPARATOR)
|
