mirror of
https://github.com/explosion/spaCy.git
synced 2024-12-27 02:16:32 +03:00
26446aa728
Move exceptions loading behind a get_tokenizer_exceptions() function for French, instead of loading into the top-level namespace. This cuts import times from 0.6s to 0.2s, at the expense of making the French data a little different from the others (there's no top-level TOKENIZER_EXCEPTIONS variable.) The current solution feels somewhat unsatisfying.
30 lines
784 B
Python
30 lines
784 B
Python
# encoding: utf8
|
|
from __future__ import unicode_literals, print_function
|
|
|
|
from ..language import Language, BaseDefaults
|
|
from ..attrs import LANG
|
|
|
|
from .language_data import *
|
|
from .punctuation import TOKENIZER_INFIXES, TOKENIZER_SUFFIXES
|
|
|
|
|
|
class FrenchDefaults(BaseDefaults):
|
|
lex_attr_getters = dict(Language.Defaults.lex_attr_getters)
|
|
lex_attr_getters[LANG] = lambda text: 'fr'
|
|
|
|
stop_words = STOP_WORDS
|
|
infixes = tuple(TOKENIZER_INFIXES)
|
|
suffixes = tuple(TOKENIZER_SUFFIXES)
|
|
token_match = TOKEN_MATCH
|
|
|
|
@classmethod
|
|
def create_tokenizer(cls, nlp=None):
|
|
cls.tokenizer_exceptions = get_tokenizer_exceptions()
|
|
return super(FrenchDefaults, cls).create_tokenizer(nlp)
|
|
|
|
|
|
class French(Language):
|
|
lang = 'fr'
|
|
|
|
Defaults = FrenchDefaults
|