2014-09-01 21:25:28 +00:00
|
|
|
# -*- coding: utf8 -*-
|
2014-08-30 18:36:06 +00:00
|
|
|
from __future__ import unicode_literals
|
2014-09-01 21:25:28 +00:00
|
|
|
import unicodedata
|
2014-10-09 21:11:31 +00:00
|
|
|
from unidecode import unidecode
|
2014-09-01 21:25:28 +00:00
|
|
|
|
|
|
|
import math
|
2014-08-30 18:36:06 +00:00
|
|
|
|
2014-09-10 16:11:13 +00:00
|
|
|
|
|
|
|
TAGS = 'adj adp adv conj det noun num pdt pos pron prt punct verb'.upper().split()
|
|
|
|
|
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
# Binary string features
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_alpha(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
return string.isalpha()
|
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_digit(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
return string.isdigit()
|
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_punct(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
for c in string:
|
2014-09-01 21:41:31 +00:00
|
|
|
if not unicodedata.category(c).startswith('P'):
|
|
|
|
return False
|
2014-09-01 21:25:28 +00:00
|
|
|
else:
|
2014-09-01 21:41:31 +00:00
|
|
|
return True
|
2014-09-01 21:25:28 +00:00
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_space(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
return string.isspace()
|
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_ascii(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
for c in string:
|
2014-09-01 21:41:31 +00:00
|
|
|
if ord(c) >= 128:
|
2014-09-01 21:25:28 +00:00
|
|
|
return False
|
|
|
|
else:
|
|
|
|
return True
|
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_title(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
return string.istitle()
|
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_lower(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
return string.islower()
|
|
|
|
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-10-09 21:11:31 +00:00
|
|
|
def is_upper(string):
|
2014-09-01 21:25:28 +00:00
|
|
|
return string.isupper()
|
2014-08-30 17:01:00 +00:00
|
|
|
|
|
|
|
|
|
|
|
# Statistics features
|
|
|
|
def oft_case(name, thresh):
|
|
|
|
def wrapped(string, prob, case_stats, tag_stats):
|
|
|
|
return string
|
|
|
|
return wrapped
|
|
|
|
|
|
|
|
|
2014-09-10 16:27:44 +00:00
|
|
|
def can_tag(name, thresh=0.5):
|
2014-08-30 17:01:00 +00:00
|
|
|
def wrapped(string, prob, case_stats, tag_stats):
|
|
|
|
return string
|
|
|
|
return wrapped
|
|
|
|
|
|
|
|
|
|
|
|
# String features
|
2014-10-29 12:19:38 +00:00
|
|
|
def canon_case(string, upper_pc=0.0, title_pc=0.0, lower_pc=0.0):
|
2014-08-30 18:57:43 +00:00
|
|
|
if upper_pc >= lower_pc and upper_pc >= title_pc:
|
|
|
|
return string.upper()
|
|
|
|
elif title_pc >= lower_pc:
|
|
|
|
return string.title()
|
|
|
|
else:
|
|
|
|
return string.lower()
|
2014-08-30 17:01:00 +00:00
|
|
|
|
2014-08-30 18:36:06 +00:00
|
|
|
|
2014-10-29 12:19:38 +00:00
|
|
|
def word_shape(string):
|
2014-08-30 17:01:00 +00:00
|
|
|
length = len(string)
|
2014-10-14 04:47:06 +00:00
|
|
|
shape = []
|
2014-08-30 17:01:00 +00:00
|
|
|
last = ""
|
|
|
|
shape_char = ""
|
|
|
|
seq = 0
|
|
|
|
for c in string:
|
|
|
|
if c.isalpha():
|
|
|
|
if c.isupper():
|
|
|
|
shape_char = "X"
|
|
|
|
else:
|
|
|
|
shape_char = "x"
|
|
|
|
elif c.isdigit():
|
|
|
|
shape_char = "d"
|
|
|
|
else:
|
|
|
|
shape_char = c
|
|
|
|
if shape_char == last:
|
|
|
|
seq += 1
|
|
|
|
else:
|
|
|
|
seq = 0
|
|
|
|
last = shape_char
|
2014-09-01 21:25:28 +00:00
|
|
|
if seq < 5:
|
2014-10-14 04:47:06 +00:00
|
|
|
shape.append(shape_char)
|
|
|
|
return ''.join(shape)
|
2014-08-30 17:01:00 +00:00
|
|
|
|
|
|
|
|
2014-10-29 12:19:38 +00:00
|
|
|
def non_sparse(string, prob, cluster, upper_pc, title_pc, lower_pc):
|
2014-10-09 21:11:31 +00:00
|
|
|
if is_alpha(string):
|
2014-10-29 12:19:38 +00:00
|
|
|
return canon_case(string, upper_pc, title_pc, lower_pc)
|
2014-09-01 21:25:28 +00:00
|
|
|
elif prob >= math.log(0.0001):
|
|
|
|
return string
|
|
|
|
else:
|
2014-10-29 12:19:38 +00:00
|
|
|
return word_shape(string)
|
2014-09-01 21:25:28 +00:00
|
|
|
|
|
|
|
|
2014-10-29 12:19:38 +00:00
|
|
|
def asciied(string):
|
2014-10-09 21:11:31 +00:00
|
|
|
ascii_string = unidecode(string)
|
|
|
|
return ascii_string.decode('ascii')
|