forked from mirrors/LibreTranslate
Remove Polyglot transliteration library
The model server for the Polyglot transliteration library is currently down and the project looks unmaintained; this currently is breaking LibreTranslate installs. This commit removes Polyglot for transliteration but keeps using Polyglot for language detection. - https://github.com/LibreTranslate/LibreTranslate/issues/344 - https://community.libretranslate.com/t/improving-transliteration-in-libretranslate/400
This commit is contained in:
parent
1c5b006c87
commit
ee638b407d
4 changed files with 5 additions and 92 deletions
|
@ -15,7 +15,7 @@ from translatehtml import translate_html
|
|||
from werkzeug.utils import secure_filename
|
||||
|
||||
from app import flood, remove_translated_files, security
|
||||
from app.language import detect_languages, transliterate, improve_translation_formatting
|
||||
from app.language import detect_languages, improve_translation_formatting
|
||||
|
||||
from .api_keys import Database, RemoteDatabase
|
||||
from .suggestions import Database as SuggestionsDatabase
|
||||
|
@ -489,8 +489,7 @@ def create_app(args):
|
|||
if text_format == "html":
|
||||
translated_text = str(translate_html(translator, text))
|
||||
else:
|
||||
translated_text = improve_translation_formatting(text, translator.translate(
|
||||
transliterate(text, target_lang=source_langs[idx]["language"])))
|
||||
translated_text = improve_translation_formatting(text, translator.translate(text))
|
||||
|
||||
results.append(unescape(translated_text))
|
||||
if source_lang == "auto":
|
||||
|
@ -512,8 +511,7 @@ def create_app(args):
|
|||
if text_format == "html":
|
||||
translated_text = str(translate_html(translator, q))
|
||||
else:
|
||||
translated_text = improve_translation_formatting(q, translator.translate(
|
||||
transliterate(q, target_lang=source_langs[0]["language"])))
|
||||
translated_text = improve_translation_formatting(q, translator.translate(q))
|
||||
|
||||
if source_lang == "auto":
|
||||
return jsonify(
|
||||
|
|
34
app/init.py
34
app/init.py
|
@ -1,6 +1,5 @@
|
|||
from pathlib import Path
|
||||
|
||||
import polyglot
|
||||
from argostranslate import package, translate
|
||||
|
||||
import app.language
|
||||
|
@ -9,7 +8,6 @@ import app.language
|
|||
def boot(load_only=None):
|
||||
try:
|
||||
check_and_install_models(load_only_lang_codes=load_only)
|
||||
check_and_install_transliteration()
|
||||
except Exception as e:
|
||||
print("Cannot update models (normal if you're offline): %s" % str(e))
|
||||
|
||||
|
@ -60,35 +58,3 @@ def check_and_install_models(force=False, load_only_lang_codes=None):
|
|||
"Loaded support for %s languages (%s models total)!"
|
||||
% (len(translate.get_installed_languages()), len(available_packages))
|
||||
)
|
||||
|
||||
|
||||
def check_and_install_transliteration(force=False):
|
||||
# 'en' is not a supported transliteration language
|
||||
transliteration_languages = [
|
||||
l.code for l in app.language.load_languages() if l.code != "en"
|
||||
]
|
||||
|
||||
# check installed
|
||||
install_needed = []
|
||||
if not force:
|
||||
t_packages_path = Path(polyglot.polyglot_path) / "transliteration2"
|
||||
for lang in transliteration_languages:
|
||||
if not (
|
||||
t_packages_path / lang / f"transliteration.{lang}.tar.bz2"
|
||||
).exists():
|
||||
install_needed.append(lang)
|
||||
else:
|
||||
install_needed = transliteration_languages
|
||||
|
||||
# install the needed transliteration packages
|
||||
if install_needed:
|
||||
print(
|
||||
f"Installing transliteration models for the following languages: {', '.join(install_needed)}"
|
||||
)
|
||||
|
||||
from polyglot.downloader import Downloader
|
||||
|
||||
downloader = Downloader()
|
||||
|
||||
for lang in install_needed:
|
||||
downloader.download(f"transliteration2.{lang}")
|
||||
|
|
|
@ -2,7 +2,6 @@ import string
|
|||
|
||||
from argostranslate import translate
|
||||
from polyglot.detect.base import Detector, UnknownLanguage
|
||||
from polyglot.transliteration.base import Transliterator
|
||||
|
||||
__languages = None
|
||||
|
||||
|
@ -113,52 +112,3 @@ def improve_translation_formatting(source, translation, improve_punctuation=True
|
|||
|
||||
return translation
|
||||
|
||||
|
||||
def __transliterate_line(transliterator, line_text):
|
||||
new_text = []
|
||||
|
||||
# transliteration is done word by word
|
||||
for orig_word in line_text.split(" "):
|
||||
# remove any punctuation on the right side
|
||||
r_word = orig_word.rstrip(string.punctuation)
|
||||
r_diff = set(char for char in orig_word) - set(char for char in r_word)
|
||||
# and on the left side
|
||||
l_word = orig_word.lstrip(string.punctuation)
|
||||
l_diff = set(char for char in orig_word) - set(char for char in l_word)
|
||||
|
||||
# the actual transliteration of the word
|
||||
t_word = transliterator.transliterate(orig_word.strip(string.punctuation))
|
||||
|
||||
# if transliteration fails, default back to the original word
|
||||
if not t_word:
|
||||
t_word = orig_word
|
||||
else:
|
||||
t_word = improve_translation_formatting(orig_word.strip(string.punctuation), t_word, improve_punctuation=False)
|
||||
|
||||
# add back any stripped punctuation
|
||||
if r_diff:
|
||||
t_word = t_word + "".join(r_diff)
|
||||
if l_diff:
|
||||
t_word = "".join(l_diff) + t_word
|
||||
|
||||
new_text.append(t_word)
|
||||
|
||||
# rebuild the text
|
||||
return " ".join(new_text)
|
||||
|
||||
|
||||
def transliterate(text, target_lang="en"):
|
||||
# initialize the transliterator from polyglot
|
||||
transliterator = Transliterator(target_lang=target_lang)
|
||||
|
||||
# check for multiline string
|
||||
if "\n" in text:
|
||||
lines = []
|
||||
# process each line separate
|
||||
for line in text.split("\n"):
|
||||
lines.append(__transliterate_line(transliterator, line))
|
||||
|
||||
# rejoin multiline string
|
||||
return "\n".join(lines)
|
||||
else:
|
||||
return __transliterate_line(transliterator, text)
|
||||
|
|
|
@ -1,6 +1,6 @@
|
|||
#!/usr/bin/env python
|
||||
import argparse
|
||||
from app.init import check_and_install_models, check_and_install_transliteration
|
||||
from app.init import check_and_install_models
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser()
|
||||
|
@ -10,4 +10,3 @@ if __name__ == "__main__":
|
|||
if len(lang_codes) == 0 or lang_codes[0] == '':
|
||||
lang_codes = None
|
||||
check_and_install_models(force=True, load_only_lang_codes=lang_codes)
|
||||
check_and_install_transliteration(force=True)
|
||||
|
|
Loading…
Reference in a new issue