def get_language(entry):
""" hacky language detection for domains. Downloads the index page, and runs
cld2 on the html """
index_url = entry.url.replace("robots.txt", "")
# hack around some issues here,
if entry.domain in KNOWN_LANGUAGES:
language = KNOWN_LANGUAGES.get(entry.domain)
else:
try:
page = requests.get(index_url)
try:
languages = cld2.detect(page.content, isPlainText=False,
hintTopLevelDomain=entry.domain.split('.')[-1])
except:
languages = cld2.detect(page.text.encode("utf8"), isPlainText=False,
hintTopLevelDomain=entry.domain.split('.')[-1])
# ignoring 'is_reliable' flag here, set on baidu.com etc (even though detects
# language appropiately
language = languages.details[0].language_name if languages.details else 'Unknown'
index_url = page.url
except Exception as e:
log.exception("Failed to analyze language for '%s'", entry.domain)
language = 'Failed'
language = language.title()
# traditional chinese -> chinese
if language == 'Chineset':
language = 'Chinese'
return language, not urlparse(index_url).netloc.endswith(entry.domain)
def do_resegment(args):
docid, text_pruned = args
lang = cld2.detect(text_pruned, want_chunks=True)
segmented = [ { "l": c[0].code,
"t": list(word_seg.segment(c[0].code, c[1])) }
for c in lang.chunks ]
return (docid, json.dumps(segmented))
def parse(self, response):
now = datetime.utcnow().replace(microsecond=0).isoformat()
url = response.url
datetimeToday = now + 'Z'
contentElement = response.css('.bpa-richtext').get()
contentElement = response.css(
'#main').get() if contentElement is None else contentElement
converter = html2text.HTML2Text()
converter.ignore_links = True
updatedDate = response.css('.bpa-time > time::attr(datetime)').get()
title = response.css('title::text').get().split(' | ')[-1]
text = converter.handle(contentElement)
isReliable, textBytesFound, details = cld2.detect(text)
language = details[0].language_name
yield {
'title': title,
'source': 'Presse- und Informationsamt der Bundesregierung',
'published': updatedDate,
'url': url,
'scraped': datetimeToday,
'classes': ['Government'],
'country': 'Germany',
'municipality': 'National',
'language': language,
'text': text
}
def parse(self, response):
now = datetime.utcnow().replace(microsecond=0).isoformat()
url = response.url
datetimeToday = now + 'Z'
textContent = 'todo'
dateElement = response.css('time::text').get()
dateElementText = dateElement.replace('\t', '').replace(
'\n', '').replace(' ',
'').replace(' ', '')
dateElementArray = dateElementText.split(',')
updatedDateISO = dateparser.parse(dateElementArray[0],
languages=['en']).date()
updatedDateTime = str(updatedDateISO)
title = response.css('h1::text').get()
contentArray = response.css('p::text').extract()
converter = html2text.HTML2Text()
converter.ignore_links = True
text = reduce(
lambda first, second: converter.handle(first) + converter.handle(
second), contentArray)
isReliable, textBytesFound, details = cld2.detect(text)
textMinusUnnecessaryChars = text.replace('\\', '')
language = details[0].language_name
yield {
'title': title,
'source': 'Georgia State Government',
'published': updatedDateTime,
'url': url,
'scraped': datetimeToday,
'classes': ['Government'],
'country': 'United States of America',
'municipality': 'Georgia',
'language': language,
'text': textMinusUnnecessaryChars
}
def main():
for fn in sys.argv[1:]:
with open(fn) as f:
guesses = cld2.detect(f.read())
sys.stdout.write("{}: {}\n"
.format(fn, " / ".join(dumplang(g)
for g in guesses)))
def detect_language(text):
# details is 3x (langName, langCode, percent, score)
lang_is_reliable, _, lang_details = cld2.detect(text)
lang_details = lang_details[0] # take only the first lang detected
lang_name, lang_code, lang_percent, lang_score = lang_details
return lang_name, lang_code, lang_score, lang_is_reliable
def do_content_extraction(args):
docid, page, baseurl = args
try:
page = zlib.decompress(page)
except:
page = ''
extr = html_extractor.ExtractedContent(baseurl, page)
lang = cld2.detect(extr.text_pruned, want_chunks=True)
segmented = [ { "l": c[0].code,
"t": list(word_seg.segment(c[0].code, c[1])) }
for c in lang.chunks ]
pagelen = len(page)
content = extr.text_content.encode("utf-8")
chash = hashlib.sha256(content).digest()
pruned = extr.text_pruned.encode("utf-8")
phash = hashlib.sha256(pruned).digest()
segmtd = json.dumps(segmented).encode("utf-8")
heads = json.dumps(extr.headings).encode("utf-8")
hhash = hashlib.sha256(heads).digest()
links = json.dumps(extr.links).encode("utf-8")
lhash = hashlib.sha256(links).digest()
rsrcs = json.dumps(extr.resources).encode("utf-8")
rhash = hashlib.sha256(rsrcs).digest()
domst = json.dumps(extr.dom_stats.to_json()).encode("utf-8")
dhash = hashlib.sha256(domst).digest()
return (docid, pagelen,
chash, content,
phash, pruned, segmtd,
hhash, heads,
lhash, links,
rhash, rsrcs,
dhash, domst)
def do_content_extraction(args):
page, url, locale, sources, access_time, result, detail, ourl, rurl = args
page = zlib.decompress(page)
pagelen = len(page)
pagehash = hashlib.sha256(page).digest()
extr = html_extractor.ExtractedContent(url, page)
langs = cld2.detect(extr.text_pruned)
return (zlib.compress(extr.text_pruned.encode("utf-8")),
zlib.compress(extr.text_content.encode("utf-8")),
zlib.compress(json.dumps(extr.headings).encode("utf-8")),
zlib.compress(json.dumps(extr.links).encode("utf-8")),
zlib.compress(json.dumps(extr.resources).encode("utf-8")),
zlib.compress(json.dumps(extr.dom_stats.to_json())
.encode("utf-8")),
langs[0].code,
langs[0].percent,
locale,
sources,
access_time,
result,
detail,
ourl,
rurl,
pagelen,
pagehash)
def check_message(message_content, pattern, set_lang="en"):
lang_details = cld2.detect(message_content)
lang = (lang_details[2][0][1])
if lang == set_lang or lang == "un":
if re.search(pattern, message_content):
return lang
else:
return None
def detect_language(text, proba_threshold):
_, _, details = cld2.detect(text)
language_code = details[0].language_code
probability = details[0].percent
if language_code != 'un' and probability > proba_threshold:
return language_code
def do_segmentation(args):
id, text = args
lang = cld2.detect(text, want_chunks=True)
segmented = [{
"l": c[0].code,
"t": list(word_seg.segment(c[0].code, c[1]))
} for c in lang.chunks]
return id, quote_utf8_as_text(json.dumps(segmented).encode("utf-8"))
def is_english(document):
'''checks if document is in English
'''
import cld2
reliable, _, details = cld2.detect(document, bestEffort=True)
return reliable and details[0][0] == 'ENGLISH' and details[0][2] >= MINENGPER
def detect_language(text):
"""Detect language using CLD2 library."""
try: # todo: figure out what's causing an occasional error
_, _, details = cld2.detect(text)
lang = details[0].language_code
except ValueError:
lang = 'un'
return lang if lang != 'un' else None
def cld2_(text, label):
try:
result = cld2.detect(text.strip())
if result[2][0].language_code == label:
return True
except Exception as e:
print(e)
pass
return False
def do_redetect(args):
id, text = args
try:
text = zlib.decompress(text)
except:
text = ''
langs = cld2.detect(text)
return (id, json.dumps([{"l":l.code, "s":l.score} for l in langs.scores]))
def guess_lang_from_data(data, is_html, default_lang='en'):
data = TextSanitizer.clean_utf8(data) # cld2 needs clean input
reliable, text_bytes, detected_languages = cld2.detect(
data.encode('utf-8', 'ignore'), isPlainText=(not is_html),
useFullLangTables=True, bestEffort=True)
if not reliable:
return default_lang
else:
return detected_languages[0][1]
def fortnite_entries(file, entries, lang, limit):
with open(file, 'r') as tsv_file:
tsv_reader = csv.reader(tsv_file, delimiter='\t')
for line in tsv_reader:
lang_details = cld2.detect(line[2])
if lang_details[2][0][1] == lang:
entries.append(Entries.EntryFortnite(line[0], line[1], line[2], line[3], lang_details[2][0][1]))
if len(entries) > limit:
break
def guess_lang_from_data(data, is_html, default_lang='en'):
assert isinstance(data, unicode)
data = TextSanitizer.clean_utf8(data) # cld2 needs clean input
reliable, text_bytes, detected_languages = cld2.detect(
data.encode('utf-8', 'ignore'), isPlainText=(not is_html),
useFullLangTables=True, bestEffort=True)
if not reliable:
return default_lang
else:
return detected_languages[0][1]
def gen_sentences(fileobj):
for line in fileobj:
line = line.strip()
if len(line.split(' ')) < 5 and len(line.encode('utf-8')) < 20:
print >> sys.stderr, "skipping too-short line"
continue
if cld2.detect(line.encode('utf-8'))[2][0][0] == 'ENGLISH':
print >> sys.stderr, "skipping English line"
continue
for sentence in tokenizer.tokenize(line):
yield sentence
def predict(self, text):
"""
Predicting the language of a text.
Parameters
----------
text : str
"""
is_reliable, text_bytes_found, details = cld2.detect(text,
bestEffort=True)
return self.map2wili(details[0].language_code)
def lang_detect(text):
printable_str = ''.join(x for x in text if x.isprintable())
isReliable, textBytesFound, details = cld2.detect(printable_str)
language_code = details[0].language_code
if language_code != 'en' and language_code != 'ru':
res = 0
elif language_code == 'ru':
res = 2
elif language_code == 'en':
res = 1
return res
def _guess_lang_from_data(data, is_html, default_lang='en'):
assert isinstance(data, unicode)
data = TextSanitizer.clean_utf8(data)
# data = data.encode('utf-8', 'ignore').decode('utf-8', 'ignore')
# print "broken", data.encode("utf-8")[17929 - 17933]
reliable, text_bytes, detected_languages = cld2.detect(
data.encode('utf-8', 'ignore'), isPlainText=(not is_html),
useFullLangTables=True, bestEffort=True)
if not reliable:
return default_lang
else:
return detected_languages[0][1]
def getlanguageSpans(documentText):
# Mimics what JANUS will do when detecting languages
# Provides a span and it's language
spanList = []
isReliable, textBytesFound, details, langVector = cld2.detect(
documentText, returnVectors=True)
for offset, num_bytes, lang_name, lang_code in langVector:
textSpan = documentText[offset:offset + num_bytes]
spanList.append((lang_code, textSpan))
return spanList
def check_language(text, languages):
"""Checks if text is written in any of the specified languages."""
import cld2
try:
_, _, lang = cld2.detect(text)
return lang[0].language_code in languages
except Exception as cld_ex: # noqa
# cld2 cannot handle some UTF-8 characters that Python can. See
# https://github.com/mikemccand/chromium-compact-language-detector/issues/22
# There is a workaround, but I'd rather just call langid in this case
import langid
lang, _ = langid.classify(text)
return lang in languages
def get_language(entry):
""" hacky language detection for domains. Downloads the index page, and runs
cld2 on the html """
index_url = entry.url.replace("robots.txt", "")
# hack around some issues here,
if entry.domain in KNOWN_LANGUAGES:
language = KNOWN_LANGUAGES.get(entry.domain)
else:
try:
page = requests.get(index_url)
try:
languages = cld2.detect(
page.content,
isPlainText=False,
hintTopLevelDomain=entry.domain.split('.')[-1])
except:
languages = cld2.detect(
page.text.encode("utf8"),
isPlainText=False,
hintTopLevelDomain=entry.domain.split('.')[-1])
# ignoring 'is_reliable' flag here, set on baidu.com etc (even though detects
# language appropiately
language = languages.details[
0].language_name if languages.details else 'Unknown'
index_url = page.url
except Exception as e:
log.exception("Failed to analyze language for '%s'", entry.domain)
language = 'Failed'
language = language.title()
# traditional chinese -> chinese
if language == 'Chineset':
language = 'Chinese'
return language, not urlparse(index_url).netloc.endswith(entry.domain)
def main():
wrapper = textwrap.TextWrapper(initial_indent=" ",
subsequent_indent=" ")
for fn in sys.argv[1:]:
with open(fn) as f:
result = cld2.detect(f.read(), want_chunks=True)
sys.stdout.write("{}: {}\n"
.format(fn, " / ".join(dumplang(g)
for g in result.scores)))
for i, chunk in enumerate(result.chunks):
sys.stdout.write(" Chunk {}: {}={}\n"
.format(i+1, chunk[0].code, chunk[0].name))
sys.stdout.write(wrapper.fill(chunk[1][:2000]))
sys.stdout.write("\n\n")
def _guess_lang_from_data(data, is_html, default_lang='en'):
assert isinstance(data, unicode)
data = TextSanitizer.clean_utf8(data)
# data = data.encode('utf-8', 'ignore').decode('utf-8', 'ignore')
# print "broken", data.encode("utf-8")[17929 - 17933]
reliable, text_bytes, detected_languages = cld2.detect(
data.encode('utf-8', 'ignore'),
isPlainText=(not is_html),
useFullLangTables=True,
bestEffort=True)
if not reliable:
return default_lang
else:
return detected_languages[0][1]
def do_content_extraction(args):
page, url, locale, sources, access_time, result, detail, ourl, rurl = args
page = zlib.decompress(page)
pagelen = len(page)
pagehash = hashlib.sha256(page).digest()
extr = html_extractor.ExtractedContent(url, page)
langs = cld2.detect(extr.text_pruned)
return (zlib.compress(extr.text_pruned.encode("utf-8")),
zlib.compress(extr.text_content.encode("utf-8")),
zlib.compress(json.dumps(extr.headings).encode("utf-8")),
zlib.compress(json.dumps(extr.links).encode("utf-8")),
zlib.compress(json.dumps(extr.resources).encode("utf-8")),
zlib.compress(
json.dumps(extr.dom_stats.to_json()).encode("utf-8")),
langs[0].code, langs[0].percent, locale, sources, access_time,
result, detail, ourl, rurl, pagelen, pagehash)
def detect(text):
text = illegalChars.sub('', text)
text = text.encode('utf-8')
try:
isReliable, textBytesFound, details = cld2.detect(text)
except:
logging.exception('Error processing text: %r', text)
return None
if isReliable:
# top language, get the language code:
# details: (('ENGLISH', 'en', 95, 1736.0), ('Unknown', 'un', 0, 0.0), ('Unknown', 'un', 0, 0.0))
return details[0][1]
else:
return None
def language(self):
"""
Detect the language of a text if no language was provided along with the text
>>> doc = Doc('Test sentence for testing text', language='en')
>>> doc.language
'en'
>>> doc = Doc('Test sentence for testing text')
>>> doc.language
'en'
"""
if not self._language:
_, _, best_guesses = cld2.detect(self.clean_text, hintLanguage=self._hint_language,
bestEffort=True)
self._language = best_guesses[0][1]
return self._language
def translate_line_or_not(line, src_lang, translator): # -> (translation, tag)
if (not line or
line.startswith('<') and line.endswith('>')):
return line, '<>'
try:
isReliable, _, lang_details = cld2.detect(line)
except ValueError:
return line, 'ValueError'
if not isReliable or lang_details[0][1].decode() != src_lang:
return line, lang_details[0][1].decode()
trans_line = translator(line)
return trans_line, src_lang
def detect(text):
text = illegalChars.sub('', text)
text = text.encode('utf-8')
try:
isReliable, textBytesFound, details = cld2.detect(text)
except:
logging.exception('Error processing text: %r', text)
return None
if isReliable:
# top language, get the language code:
# details: (('ENGLISH', 'en', 95, 1736.0), ('Unknown', 'un', 0, 0.0), ('Unknown', 'un', 0, 0.0))
return details[0][1]
else:
return None
def cld2_(text, label):
score = 0
try:
result = cld2.detect(text.strip())
if result[2][0].language_code in label:
score += 1
if result[2][1].language_code in label:
score += 1
if result[2][0].language_code == label[0]:
score += 2
if result[2][0].language_code == label[0] and result[2][
1].language_code == label[1]:
score += 3
except Exception as e:
print(e)
pass
return score
def detect_tweet_lang(text):
try:
#Remove mentions and URLs before trying to detect language
text = RE_MENTION.sub(" ", text)
text = RE_URL.sub(" ", text)
vals = {}
text = text.encode("UTF-8")
vals["cld_reliable"], vals["cld_bytes"], details = cld2.detect(text)
if len(details) > 1:
vals["cld_lang1"] = details[0][1]
vals["cld_lang1_percent"] = details[0][2]
if len(details) > 2:
vals["cld_lang2"] = details[1][1]
vals["cld_lang2_percent"] = details[1][2]
return vals
except Exception as e2:
logging.debug("CLD error: " + str(e2) + "\n" + text)
def detect_from_document(self, document, html=None):
return cld2.detect(
html or document.source_data,
isPlainText=False,
useFullLangTables=True,
hintTopLevelDomain=document.get_url().
suffix, # TODO: doesn't seem to have any influence?
hintLanguage=None,
hintLanguageHTTPHeaders=None, # TODO from headers
hintEncoding=None, # TODO from headers
returnVectors=False,
debugScoreAsQuads=False,
debugHTML=False,
bestEffort=True,
debugCR=False,
debugVerbose=False,
debugQuiet=True,
debugEcho=False)
def detect_from_document(self, document, html=None):
return cld2.detect(
html or document.source_data,
isPlainText=False,
useFullLangTables=True,
hintTopLevelDomain=document.get_url().suffix, # TODO: doesn't seem to have any influence?
hintLanguage=None,
hintLanguageHTTPHeaders=None, # TODO from headers
hintEncoding=None, # TODO from headers
returnVectors=False,
debugScoreAsQuads=False,
debugHTML=False,
bestEffort=True,
debugCR=False,
debugVerbose=False,
debugQuiet=True,
debugEcho=False
)
def do_content_extraction(args):
origin, page, baseurl = args
try:
page = zlib.decompress(page)
except:
page = ''
pagelen = len(page)
extr = html_extractor.ExtractedContent(baseurl, page)
langs = cld2.detect(extr.text_pruned)
pcontent = zlib.compress(extr.text_pruned.encode("utf-8"))
phash = hashlib.sha256(pcontent).digest()
headings = zlib.compress(json.dumps(extr.headings).encode("utf-8"))
links = zlib.compress(json.dumps(extr.links).encode("utf-8"))
resources = zlib.compress(json.dumps(extr.resources).encode("utf-8"))
domstats = zlib.compress(json.dumps(extr.dom_stats.to_json()).encode("utf-8"))
return (origin, pagelen, phash, langs[0].code, langs[0].percent,
pcontent, links, resources, headings, domstats)
def do_content_extraction(args):
origin, page, baseurl = args
try:
page = zlib.decompress(page)
except:
page = ''
pagelen = len(page)
extr = html_extractor.ExtractedContent(baseurl, page)
langs = cld2.detect(extr.text_pruned)
pcontent = zlib.compress(extr.text_pruned.encode("utf-8"))
phash = hashlib.sha256(pcontent).digest()
headings = zlib.compress(json.dumps(extr.headings).encode("utf-8"))
links = zlib.compress(json.dumps(extr.links).encode("utf-8"))
resources = zlib.compress(json.dumps(extr.resources).encode("utf-8"))
domstats = zlib.compress(
json.dumps(extr.dom_stats.to_json()).encode("utf-8"))
return (origin, pagelen, phash, langs[0].code, langs[0].percent, pcontent,
links, resources, headings, domstats)
def parse(self, response):
now = datetime.utcnow().replace(microsecond=0).isoformat()
url = response.url
datetimeToday = now + 'Z'
textContent = 'todo'
dateElement = response.css('time::text').get()
dateElementText = dateElement.replace('\t', '').replace(
'\n', '').replace(' ',
'').replace(' ', '')
dateElementArray = dateElementText.split(',')
updatedDateISO = dateparser.parse(dateElementArray[0],
languages=['en']).date()
updatedDateTime = str(updatedDateISO)
tempTitle = response.css('.content-title.text-center::text').get()
if (tempTitle == '\n '):
title = response.css(".field--item::text").get()
else:
title = tempTitle
contentArray = response.css('p::text').extract()
converter = html2text.HTML2Text()
converter.ignore_links = True
text = reduce(
lambda first, second: converter.handle(first) + converter.handle(
second), contentArray)
isReliable, textBytesFound, details = cld2.detect(text)
textMinusUnnecessaryChars = text.replace(
"Federal government websites often end in .gov or .mil. Before sharing\nsensitive information, make sure you're on a federal government site. The\nensures that you are connecting to the official website and that any\ninformation you provide is encrypted and transmitted securely.",
"").replace('\\', '')
language = details[0].language_name
yield {
'title': title,
'source': 'Federal Dept. Of Agriculture',
'published': updatedDateISO,
'url': url,
'scraped': datetimeToday,
'classes': ['Government'],
'country': 'United States of America',
'municipality': 'Federal Government',
'language': language,
'text': textMinusUnnecessaryChars
}
def should_be_good_language(self, lang_interface="fil"):
lang_interface = MainPageLocators.LANGUAGE_BUTTONS_DICT[
lang_interface][1]
text_on_page = ""
for i in range(1, 6):
text_on_page = text_on_page + WebDriverWait(self.browser, 5).until(
EC.presence_of_element_located(
MainPageLocators.HEADER_DICT[i])).text + " "
title = WebDriverWait(self.browser, 5).until(
EC.visibility_of_element_located(MainPageLocators.TITLE_TEXT)).text
text_on_page += title
for i in range(1, 9):
text_on_page = text_on_page + " " + WebDriverWait(
self.browser, 5).until(
EC.presence_of_element_located(
MainPageLocators.COLUMNS_DICT[i])).text
isReliable, textBytesFound, lang_lit = cld2.detect(text_on_page)
detect_lang = lang_lit[0][1]
assert detect_lang == lang_interface, f"The title was not changed, detected {detect_lang}, used {lang_interface}"
def cld2_detector(text):
"""
cld2 library functionality implementation for the language detection API
this library is used for multiple language detection
Args:
text: (string) text value sent for language detection
Returns: (dict) detected languages with its confidence values or (boolean) if error returns false
"""
try:
text = ''.join(x for x in text if x in string.printable)
result = cld2.detect(text.strip())
if result[2][1].language_code != "un":
return [(result[2][0].language_code, result[2][0].percent),
(result[2][1].language_code, result[2][1].percent)]
else:
return [(result[2][0].language_code, result[2][0].percent)]
except Exception as e:
return str(e)
def language_code_for_text(text: str):
"""Returns an ISO 690 language code for the plain text passed as a parameter.
:param text: Text that should be identified
:return: ISO 690 language code (e.g. 'en') on successful identification, empty string ('') on failure
"""
text = decode_object_from_bytes_if_needed(text)
if not text:
return ''
if len(text) > __MAX_TEXT_LENGTH:
log.warning("Text is longer than %d, trimming..." % __MAX_TEXT_LENGTH)
text = text[:__MAX_TEXT_LENGTH]
# We need to verify that the file can cleany encode and decode because CLD can segfault on bad UTF-8
text = __recode_utf8_string(text)
try:
is_reliable, text_bytes_found, details = cld2.detect(
utf8Bytes=text, useFullLangTables=True)
except Exception as ex:
log.error("Error while detecting language: %s" % str(ex))
return ''
if not details:
return ''
best_match = details[0]
language_name = best_match.language_name.lower()
language_code = best_match.language_code.lower()
if language_name in {'unknown', 'tg_unknown_language'
} or language_code == 'un':
return ''
if not language_is_supported(language_code):
return ''
return language_code
def language_code_for_text(text: str):
"""Returns an ISO 690 language code for the plain text passed as a parameter.
:param text: Text that should be identified
:return: ISO 690 language code (e.g. 'en') on successful identification, empty string ('') on failure
"""
text = decode_object_from_bytes_if_needed(text)
if not text:
return ''
if len(text) > __MAX_TEXT_LENGTH:
log.warning("Text is longer than %d, trimming..." % __MAX_TEXT_LENGTH)
text = text[:__MAX_TEXT_LENGTH]
# We need to verify that the file can cleany encode and decode because CLD can segfault on bad UTF-8
text = __recode_utf8_string(text)
try:
is_reliable, text_bytes_found, details = cld2.detect(utf8Bytes=text, useFullLangTables=True)
except Exception as ex:
log.error("Error while detecting language: %s" % str(ex))
return ''
if not details:
return ''
best_match = details[0]
language_name = best_match.language_name.lower()
language_code = best_match.language_code.lower()
if language_name in {'unknown', 'tg_unknown_language'} or language_code == 'un':
return ''
if not language_is_supported(language_code):
return ''
return language_code
def is_slovene(content): a, b, languages = cld2.detect(content) return 'SLOVENIAN' in [x[0] for x in languages]