mirror of
https://github.com/Helsinki-NLP/OPUS-MT-train.git
synced 2024-09-17 15:17:22 +03:00
183 lines
7.5 KiB
Makefile
183 lines
7.5 KiB
Makefile
|
|
LOAD_MODULES = module use -a /projappl/nlpl/software/modules/etc/ && \
|
|
module load nlpl-udpipe nlpl-opus &&
|
|
|
|
WIKILANGS = ${sort $(patsubst >%wiki-,%,${shell grep -o '>..wiki-' index.html})} \
|
|
${sort $(patsubst >%wiki-,%,${shell grep -o '>...wiki-' index.html})}
|
|
|
|
|
|
LANGID = af
|
|
WIKI_TXT = wiki.${LANGID}.gz
|
|
|
|
LANGNAME = ${shell ${LOAD_MODULES} opus-iso639 -e ${LANGID} | \
|
|
cut -f1 -d';' | tr ' ' '-' | tr '[:upper:]' '[:lower:]'}
|
|
UDPIPE_MODEL = ${notdir $(shell find ${UDPIPE_MODELS}/ -name "${LANGNAME}*.udpipe" | head -1)}
|
|
WIKI_JSON = $(shell grep -o '${LANGID}wiki-[0-9]*-cirrussearch-content.json.gz' index.html | head -1)
|
|
|
|
|
|
all: index.html
|
|
for l in ${WIKILANGS}; do \
|
|
${MAKE} LANGID=$$l wiki-txt; \
|
|
done
|
|
|
|
wiki-txt:
|
|
if [ "${UDPIPE_MODEL}" != "" ]; then \
|
|
${MAKE} ${WIKI_TXT}; \
|
|
fi
|
|
|
|
print-names:
|
|
echo ${LANGNAME}
|
|
echo ${UDPIPE_MODEL}
|
|
echo ${WIKI_JSON}
|
|
|
|
|
|
${WIKI_JSON}:
|
|
wget https://dumps.wikimedia.org/other/cirrussearch/current/${WIKI_JSON}
|
|
|
|
|
|
${WIKI_TXT}: ${WIKI_JSON}
|
|
${LOAD_MODULES} \
|
|
zcat $< | jq -r '.text' | \
|
|
grep -v 'null' |\
|
|
udpipe --input=horizontal --tokenize \
|
|
${UDPIPE_MODELS}/${UDPIPE_MODEL} |\
|
|
grep '^\# *text *= *' |\
|
|
sed 's/^\# *text *= *//' |\
|
|
gzip -c > $@
|
|
|
|
index.html:
|
|
wget -O $@ https://dumps.wikimedia.org/other/cirrussearch/current
|
|
|
|
|
|
|
|
|
|
|
|
# afrikaans-afribooms-ud-2.4-190531.udpipe af
|
|
# ancient_greek-perseus-ud-2.4-190531.udpipe
|
|
# ancient_greek-proiel-ud-2.4-190531.udpipe
|
|
# arabic-padt-ud-2.4-190531.udpipe ar
|
|
# armenian-armtdp-ud-2.4-190531.udpipe
|
|
# basque-bdt-ud-2.4-190531.udpipe eo
|
|
# belarusian-hse-ud-2.4-190531.udpipe
|
|
# bulgarian-btb-ud-2.4-190531.udpipe bg
|
|
# catalan-ancora-ud-2.4-190531.udpipe ca
|
|
# chinese-gsd-ud-2.4-190531.udpipe zh
|
|
# classical_chinese-kyoto-ud-2.4-190531.udpipe zh_tw
|
|
# coptic-scriptorium-ud-2.4-190531.udpipe
|
|
# croatian-set-ud-2.4-190531.udpipe hr
|
|
# czech-cac-ud-2.4-190531.udpipe cs
|
|
# czech-cltt-ud-2.4-190531.udpipe cs
|
|
# czech-fictree-ud-2.4-190531.udpipe cs
|
|
# czech-pdt-ud-2.4-190531.udpipe cs
|
|
# danish-ddt-ud-2.4-190531.udpipe da
|
|
# dutch-alpino-ud-2.4-190531.udpipe nl
|
|
# dutch-lassysmall-ud-2.4-190531.udpipe nl
|
|
# english-ewt-ud-2.4-190531.udpipe en
|
|
# english-gum-ud-2.4-190531.udpipe en
|
|
# english-lines-ud-2.4-190531.udpipe en
|
|
# english-partut-ud-2.4-190531.udpipe en
|
|
# estonian-edt-ud-2.4-190531.udpipe et
|
|
# estonian-ewt-ud-2.4-190531.udpipe et
|
|
# finnish-ftb-ud-2.4-190531.udpipe fi
|
|
# finnish-tdt-ud-2.4-190531.udpipe fi
|
|
# french-gsd-ud-2.4-190531.udpipe fr
|
|
# french-partut-ud-2.4-190531.udpipe fr
|
|
# french-sequoia-ud-2.4-190531.udpipe fr
|
|
# french-spoken-ud-2.4-190531.udpipe fr
|
|
# galician-ctg-ud-2.4-190531.udpipe gl
|
|
# galician-treegal-ud-2.4-190531.udpipe gl
|
|
# german-gsd-ud-2.4-190531.udpipe de
|
|
# gothic-proiel-ud-2.4-190531.udpipe
|
|
# greek-gdt-ud-2.4-190531.udpipe el
|
|
# hebrew-htb-ud-2.4-190531.udpipe he
|
|
# hindi-hdtb-ud-2.4-190531.udpipe hi
|
|
# hungarian-szeged-ud-2.4-190531.udpipe hu
|
|
# indonesian-gsd-ud-2.4-190531.udpipe id
|
|
# irish-idt-ud-2.4-190531.udpipe cy
|
|
# italian-isdt-ud-2.4-190531.udpipe it
|
|
# italian-partut-ud-2.4-190531.udpipe it
|
|
# italian-postwita-ud-2.4-190531.udpipe it
|
|
# italian-vit-ud-2.4-190531.udpipe it
|
|
# japanese-gsd-ud-2.4-190531.udpipe ja
|
|
# korean-gsd-ud-2.4-190531.udpipe ko
|
|
# korean-kaist-ud-2.4-190531.udpipe ko
|
|
# latin-ittb-ud-2.4-190531.udpipe la
|
|
# latin-perseus-ud-2.4-190531.udpipe la
|
|
# latin-proiel-ud-2.4-190531.udpipe la
|
|
# latvian-lvtb-ud-2.4-190531.udpipe lv
|
|
# lithuanian-alksnis-ud-2.4-190531.udpipe lt
|
|
# lithuanian-hse-ud-2.4-190531.udpipe lt
|
|
# maltese-mudt-ud-2.4-190531.udpipe mt
|
|
# marathi-ufal-ud-2.4-190531.udpipe
|
|
# north_sami-giella-ud-2.4-190531.udpipe
|
|
# norwegian-bokmaal-ud-2.4-190531.udpipe nb
|
|
# norwegian-nynorsklia-ud-2.4-190531.udpipe nn
|
|
# norwegian-nynorsk-ud-2.4-190531.udpipe nn
|
|
# old_church_slavonic-proiel-ud-2.4-190531.udpipe
|
|
# old_french-srcmf-ud-2.4-190531.udpipe
|
|
# old_russian-torot-ud-2.4-190531.udpipe
|
|
# persian-seraji-ud-2.4-190531.udpipe fa
|
|
# polish-lfg-ud-2.4-190531.udpipe pl
|
|
# polish-pdb-ud-2.4-190531.udpipe pl
|
|
# portuguese-bosque-ud-2.4-190531.udpipe pt
|
|
# portuguese-gsd-ud-2.4-190531.udpipe pt
|
|
# romanian-nonstandard-ud-2.4-190531.udpipe ro
|
|
# romanian-rrt-ud-2.4-190531.udpipe ro
|
|
# russian-gsd-ud-2.4-190531.udpipe ru
|
|
# russian-syntagrus-ud-2.4-190531.udpipe ru
|
|
# russian-taiga-ud-2.4-190531.udpipe ru
|
|
# serbian-set-ud-2.4-190531.udpipe sr
|
|
# slovak-snk-ud-2.4-190531.udpipe sk
|
|
# slovenian-ssj-ud-2.4-190531.udpipe sl
|
|
# slovenian-sst-ud-2.4-190531.udpipe sl
|
|
# spanish-ancora-ud-2.4-190531.udpipe es
|
|
# spanish-gsd-ud-2.4-190531.udpipe es
|
|
# swedish-lines-ud-2.4-190531.udpipe sv
|
|
# swedish-talbanken-ud-2.4-190531.udpipe sv
|
|
# tamil-ttb-ud-2.4-190531.udpipe
|
|
# telugu-mtg-ud-2.4-190531.udpipe
|
|
# turkish-imst-ud-2.4-190531.udpipe tr
|
|
# ukrainian-iu-ud-2.4-190531.udpipe
|
|
# urdu-udtb-ud-2.4-190531.udpipe
|
|
# uyghur-udt-ud-2.4-190531.udpipe
|
|
# vietnamese-vtb-ud-2.4-190531.udpipe vi
|
|
# wolof-wtb-ud-2.4-190531.udpipe
|
|
|
|
|
|
|
|
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwiki-20191209-cirrussearch-content.json.gz
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwiki-20191209-cirrussearch-general.json.gz
|
|
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwikibooks-20191209-cirrussearch-content.json.gz
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwikibooks-20191209-cirrussearch-general.json.gz
|
|
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwikiquote-20191209-cirrussearch-content.json.gz
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwikiquote-20191209-cirrussearch-general.json.gz
|
|
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwiktionary-20191209-cirrussearch-content.json.gz
|
|
# https://dumps.wikimedia.org/other/cirrussearch/current/afwiktionary-20191209-cirrussearch-general.json.gz
|
|
|
|
|
|
# enwiki-20191209-cirrussearch-content.json.gz 10-Dec-2019 11:04 22622822308
|
|
# enwiki-20191209-cirrussearch-content.json.gz.tmp 10-Dec-2019 10:57 21460369408
|
|
# enwiki-20191209-cirrussearch-general.json.gz 10-Dec-2019 16:22 50406322974
|
|
# enwiki-20191209-cirrussearch-general.json.gz.tmp 10-Dec-2019 15:50 44720914432
|
|
# enwikibooks-20191209-cirrussearch-content.json.gz 10-Dec-2019 16:24 319454731
|
|
# enwikibooks-20191209-cirrussearch-general.json.gz 10-Dec-2019 16:25 97206925
|
|
# enwikinews-20191209-cirrussearch-content.json.gz 10-Dec-2019 16:26 53746769
|
|
# enwikinews-20191209-cirrussearch-general.json.gz 10-Dec-2019 16:36 364098656
|
|
# enwikiquote-20191209-cirrussearch-content.json.gz 10-Dec-2019 16:38 234637326
|
|
# enwikiquote-20191209-cirrussearch-general.json.gz 10-Dec-2019 16:38 66848855
|
|
# enwikisource-20191209-cirrussearch-content.json.gz 10-Dec-2019 17:09 5236203374
|
|
# enwikisource-20191209-cirrussearch-content.json..> 10-Dec-2019 17:06 4597481472
|
|
# enwikisource-20191209-cirrussearch-general.json.gz 10-Dec-2019 17:11 152492247
|
|
# enwikiversity-20191209-cirrussearch-content.jso..> 10-Dec-2019 17:12 145288148
|
|
# enwikiversity-20191209-cirrussearch-general.jso..> 10-Dec-2019 17:13 193051475
|
|
# enwikivoyage-20191209-cirrussearch-content.json.gz 10-Dec-2019 17:14 179134384
|
|
# enwikivoyage-20191209-cirrussearch-general.json.gz 10-Dec-2019 17:15 99357806
|
|
# enwiktionary-20191209-cirrussearch-content.json.gz 10-Dec-2019 17:36 2319801836
|
|
# enwiktionary-20191209-cirrussearch-content.json..> 10-Dec-2019 17:23 918503424
|
|
# enwiktionary-20191209-cirrussearch-general.json.gz 10-Dec-2019 17:42 848846623
|
|
# enwiktionary-20191209-cirrussearch-general.json..> 10-Dec-2019 17:40 661585920
|