2023-06-26 12:41:03 +03:00
|
|
|
|
from ...symbols import NORM, ORTH
|
2021-01-27 05:29:22 +03:00
|
|
|
|
from ...util import update_exc
|
2023-06-26 12:41:03 +03:00
|
|
|
|
from ..tokenizer_exceptions import BASE_EXCEPTIONS
|
2021-01-22 20:08:41 +03:00
|
|
|
|
|
|
|
|
|
|
_exc = {}
|
|
|
|
|
|
|
|
|
|
|
|
_abbrev_exc = [
|
|
|
|
|
|
# Weekdays abbreviations
|
2021-01-27 14:11:42 +03:00
|
|
|
|
{ORTH: "дүй", NORM: "дүйшөмбү"},
|
|
|
|
|
|
{ORTH: "шей", NORM: "шейшемби"},
|
|
|
|
|
|
{ORTH: "шар", NORM: "шаршемби"},
|
|
|
|
|
|
{ORTH: "бей", NORM: "бейшемби"},
|
|
|
|
|
|
{ORTH: "жум", NORM: "жума"},
|
|
|
|
|
|
{ORTH: "ишм", NORM: "ишемби"},
|
|
|
|
|
|
{ORTH: "жек", NORM: "жекшемби"},
|
2021-01-22 20:08:41 +03:00
|
|
|
|
# Months abbreviations
|
2021-01-27 14:11:42 +03:00
|
|
|
|
{ORTH: "янв", NORM: "январь"},
|
|
|
|
|
|
{ORTH: "фев", NORM: "февраль"},
|
|
|
|
|
|
{ORTH: "мар", NORM: "март"},
|
|
|
|
|
|
{ORTH: "апр", NORM: "апрель"},
|
|
|
|
|
|
{ORTH: "июн", NORM: "июнь"},
|
|
|
|
|
|
{ORTH: "июл", NORM: "июль"},
|
|
|
|
|
|
{ORTH: "авг", NORM: "август"},
|
|
|
|
|
|
{ORTH: "сен", NORM: "сентябрь"},
|
|
|
|
|
|
{ORTH: "окт", NORM: "октябрь"},
|
|
|
|
|
|
{ORTH: "ноя", NORM: "ноябрь"},
|
|
|
|
|
|
{ORTH: "дек", NORM: "декабрь"},
|
2021-01-22 20:08:41 +03:00
|
|
|
|
# Number abbreviations
|
2021-01-27 14:11:42 +03:00
|
|
|
|
{ORTH: "млрд", NORM: "миллиард"},
|
|
|
|
|
|
{ORTH: "млн", NORM: "миллион"},
|
2021-01-22 20:08:41 +03:00
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
|
for abbr in _abbrev_exc:
|
|
|
|
|
|
for orth in (abbr[ORTH], abbr[ORTH].capitalize(), abbr[ORTH].upper()):
|
2021-01-27 14:11:42 +03:00
|
|
|
|
_exc[orth] = [{ORTH: orth, NORM: abbr[NORM]}]
|
|
|
|
|
|
_exc[orth + "."] = [{ORTH: orth + ".", NORM: abbr[NORM]}]
|
2021-01-22 20:08:41 +03:00
|
|
|
|
|
|
|
|
|
|
for exc_data in [ # "etc." abbreviations
|
|
|
|
|
|
{ORTH: "ж.б.у.с.", NORM: "жана башка ушул сыяктуу"},
|
|
|
|
|
|
{ORTH: "ж.б.", NORM: "жана башка"},
|
|
|
|
|
|
{ORTH: "ж.", NORM: "жыл"},
|
|
|
|
|
|
{ORTH: "б.з.ч.", NORM: "биздин заманга чейин"},
|
|
|
|
|
|
{ORTH: "б.з.", NORM: "биздин заман"},
|
|
|
|
|
|
{ORTH: "кк.", NORM: "кылымдар"},
|
|
|
|
|
|
{ORTH: "жж.", NORM: "жылдар"},
|
|
|
|
|
|
{ORTH: "к.", NORM: "кылым"},
|
|
|
|
|
|
{ORTH: "көч.", NORM: "көчөсү"},
|
|
|
|
|
|
{ORTH: "м-н", NORM: "менен"},
|
|
|
|
|
|
{ORTH: "б-ча", NORM: "боюнча"},
|
2021-01-30 04:52:33 +03:00
|
|
|
|
]:
|
|
|
|
|
|
_exc[exc_data[ORTH]] = [exc_data]
|
2021-01-22 20:08:41 +03:00
|
|
|
|
|
2021-01-27 05:29:22 +03:00
|
|
|
|
TOKENIZER_EXCEPTIONS = update_exc(BASE_EXCEPTIONS, _exc)
|
