TYPES_REGEX = '|'.join(
map(lambda x: '[{}{}]{}'.format(x[0], x[0].upper(), x[1:]), TYPES)
)
def _clean_query(q):
q = re.sub('c(e|é)dex ?[\d]*', '', q, flags=re.IGNORECASE)
q = re.sub('bp ?[\d]*', '', q, flags=re.IGNORECASE)
q = re.sub('cs ?[\d]*', '', q, flags=re.IGNORECASE)
q = re.sub('\d{,2}(e|[eè]me) ([eé]tage)', '', q, flags=re.IGNORECASE)
q = re.sub(' {2,}', ' ', q, flags=re.IGNORECASE)
q = re.sub('[ -]s/[ -]', ' sur ', q, flags=re.IGNORECASE)
q = re.sub('[ -]s/s[ -]', ' sous ', q, flags=re.IGNORECASE)
q = q.strip()
return q
clean_query = yielder(_clean_query)
def _extract_address(q):
m = extract_address_pattern.search(q)
return m.group() if m else q
extract_address_pattern = re.compile(
'(\d+( *(bis|ter))?,? +(' + TYPES_REGEX + ') .*(\d{5})?).*',
flags=re.IGNORECASE)
extract_address = yielder(_extract_address)
def _glue_ordinal(q):
"""Glue '3' and 'bis'."""
return glue_ordinal_pattern.sub('\g<1>\g<2>\g<3>', q)
ORDINAL_REGEX = 'bis|ter|quater|quinquies|sexies|[a-z]'
("ck", "k"),
("ph", "f"),
("th$", "te"), # This t sounds.
("(?<=[^sc0-9])h", ""),
("^h(?=.)+", ""),
("sc", "s"),
("sh", "ch"),
("((?<=[^0-9])w|^w)", "v"),
("c(?=[eiy])", "s"),
("(?<=[^0-9])y", "i"),
("esn", "en"),
("oe(?=\\w)", "e"),
("(?<=[^0-9])s$", ""),
("(?<=u)l?x$", ""), # eaux, eux, aux, aulx
("(?<=u)lt$", "t"),
("(?<=[a-z])[dg]$", ""),
("(?<=[^es0-9])t$", ""),
("(?<=[aeiou])(m)(?=[pbgf])", "n"),
("(?<=[a-z]{2})(e$)", ""), # Remove "e" at last position only if
# it follows two letters?
("(\\D)(?=\\1)", ""), # Remove duplicate letters.
)
_s = s
for pattern, repl in rules:
_s = re.sub(pattern, repl, _s)
_CACHE[s] = _s
return _CACHE[s]
phonemicize = yielder(_phonemicize)
map(lambda x: '[{}{}]{}'.format(x[0], x[0].upper(), x[1:]), TYPES))
def _clean_query(q):
q = re.sub('c(e|é)dex ?[\d]*', '', q, flags=re.IGNORECASE)
q = re.sub('bp ?[\d]*', '', q, flags=re.IGNORECASE)
q = re.sub('cs ?[\d]*', '', q, flags=re.IGNORECASE)
q = re.sub('\d{,2}(e|[eè]me) ([eé]tage)', '', q, flags=re.IGNORECASE)
q = re.sub(' {2,}', ' ', q, flags=re.IGNORECASE)
q = re.sub('[ -]s/[ -]', ' sur ', q, flags=re.IGNORECASE)
q = re.sub('[ -]s/s[ -]', ' sous ', q, flags=re.IGNORECASE)
q = q.strip()
return q
clean_query = yielder(_clean_query)
def _extract_address(q):
m = extract_address_pattern.search(q)
return m.group() if m else q
extract_address_pattern = re.compile('(\d+( *(bis|ter))?,? +(' + TYPES_REGEX +
') .*(\d{5})?).*',
flags=re.IGNORECASE)
extract_address = yielder(_extract_address)
def _glue_ordinal(q):
"""Glue '3' and 'bis'."""
from addok.utils import yielder
from . import normalize as _normalize
from . import synonymize as _synonymize
from . import tokenize as _tokenize
def tokenize(pipe):
for text in pipe:
for token in _tokenize(text):
yield token
normalize = yielder(_normalize)
synonymize = yielder(_synonymize)
("cc(?=[ie])", "s"), # Others will hit the c => k and deduplicate
("ck", "k"),
("ph", "f"),
("th$", "te"), # This t sounds.
("(?<=[^sc0-9])h", ""),
("^h(?=.)+", ""),
("sc", "s"),
("sh", "ch"),
("((?<=[^0-9])w|^w)", "v"),
("c(?=[eiy])", "s"),
("(?<=[^0-9])y", "i"),
("esn", "en"),
("oe(?=\\w)", "e"),
("(?<=[^0-9])s$", ""),
("(?<=u)l?x$", ""), # eaux, eux, aux, aulx
("(?<=u)lt$", "t"),
("(?<=[a-z])[dg]$", ""),
("(?<=[^es0-9])t$", ""),
("(?<=[aeiou])(m)(?=[pbgf])", "n"),
("(?<=[a-z]{2})(e$)", ""), # Remove "e" at last position only if
# it follows two letters?
("(\\D)(?=\\1)", ""), # Remove duplicate letters.
)
_s = s
for pattern, repl in rules:
_s = re.sub(pattern, repl, _s)
_CACHE[s] = _s
return _CACHE[s]
phonemicize = yielder(_phonemicize)
