def test_is_same_ending_us_uk_spelling():
assert is_same_ending_us_uk_spelling('centre', 'center')
assert not is_same_ending_us_uk_spelling('flower', 'flower')
assert not is_same_ending_us_uk_spelling('flower', 'bowl')
assert not is_same_ending_us_uk_spelling('ter', 'atre')
assert not is_same_ending_us_uk_spelling('er', 're')
assert not is_same_ending_us_uk_spelling('centre', 'cinter')
def validate(self):
"""Check whether or not this substitution is worth keeping."""
token1, token2 = self.tokens
lem1, lem2 = self.lemmas
tokens1, tokens2 = self.source.tokens, self.destination.tokens
lemmas1, lemmas2 = self.source.lemmas, self.destination.lemmas
# Only real-word lemmas.
wordnet_words = _get_wordnet_words()
if lem1 not in wordnet_words or lem2 not in wordnet_words:
return False
# '21st'/'twenty-first', etc.
if (is_int(token1[0]) or is_int(token2[0]) or
is_int(lem1[0]) or is_int(lem2[0])):
return False
# 'sen'/'senator', 'gov'/'governor', 'nov'/'november', etc.
if (token1 == token2[:3] or token2 == token1[:3] or
lem1 == lem2[:3] or lem2 == lem1[:3]):
return False
# 'programme'/'program', etc.
if (token1[:-2] == token2 or token2[:-2] == token1 or
lem1[:-2] == lem2 or lem2[:-2] == lem1):
return False
# 'centre'/'center', etc.
if is_same_ending_us_uk_spelling(token1, token2):
return False
if is_same_ending_us_uk_spelling(lem1, lem2):
return False
# stopwords
if (token1 in stopwords or token2 in stopwords or
lem1 in stopwords or lem2 in stopwords):
return False
# Other minor spelling changes, also catching cases where tokens are
# not different but lemmas are (because of lemmatization fluctuations).
if levenshtein(token1, token2) <= 1:
return False
if levenshtein(lem1, lem2) <= 1:
return False
# Word deletion ('high school' -> 'school')
if (self.start + self.position > 0 and
(token2 == tokens1[self.start + self.position - 1] or
lem2 == lemmas1[self.start + self.position - 1])):
return False
if (self.start + self.position < len(tokens1) - 1 and
(token2 == tokens1[self.start + self.position + 1] or
lem2 == lemmas1[self.start + self.position + 1])):
return False
# Word insertion ('school' -> 'high school')
if (self.position > 0 and
(token1 == tokens2[self.position - 1] or
lem1 == lemmas2[self.position - 1])):
return False
if (self.position < len(tokens2) - 1 and
(token1 == tokens2[self.position + 1] or
lem1 == lemmas2[self.position + 1])):
return False
# Two words deletion ('supply of energy' -> 'supply')
if (self.start + self.position > 1 and
(token2 == tokens1[self.start + self.position - 2] or
lem2 == lemmas1[self.start + self.position - 2])):
return False
if (self.start + self.position < len(tokens1) - 2 and
(token2 == tokens1[self.start + self.position + 2] or
lem2 == lemmas1[self.start + self.position + 2])):
return False
# Words stuck together ('policy maker' -> 'policymaker'
# or 'policy-maker')
if (self.start + self.position > 0 and
(token2 == tokens1[self.start + self.position - 1] + token1 or
token2 == tokens1[self.start + self.position - 1] +
'-' + token1 or
lem2 == lemmas1[self.start + self.position - 1] + lem1 or
lem2 == lemmas1[self.start + self.position - 1] + '-' + lem1)):
return False
if (self.start + self.position < len(tokens1) - 1 and
(token2 == token1 + tokens1[self.start + self.position + 1] or
token2 == token1 + '-' +
tokens1[self.start + self.position + 1] or
lem2 == lem1 + lemmas1[self.start + self.position + 1] or
lem2 == lem1 + '-' + lemmas1[self.start + self.position + 1])):
return False
# Words separated ('policymaker' or 'policy-maker' -> 'policy maker')
if (self.position > 0 and
(token1 == tokens2[self.position - 1] + token2 or
token1 == tokens2[self.position - 1] + '-' + token2 or
lem1 == lemmas2[self.position - 1] + lem2 or
lem1 == lemmas2[self.position - 1] + '-' + lem2)):
return False
if (self.position < len(tokens2) - 1 and
(token1 == token2 + tokens2[self.position + 1] or
token1 == token2 + '-' + tokens2[self.position + 1] or
lem1 == lem2 + lemmas2[self.position + 1] or
lem1 == lem2 + '-' + lemmas2[self.position + 1])):
return False
# We need 2 extra checks compare to the words-stuck-together situation,
# to detect teh second substitution appearing because of word
# separation. Indeed in this case, contrary to words-stuck-together, we
# can't rely on word shifts always being present, since the destination
# can be cut shorter. In other words, in the following case:
# (1) i'll come anytime there
# (2) i'll come any time
# these checks let us exclude 'there' -> 'time' as a substitution (in
# the words-stuck-together case, the word 'there' would be present in
# both sentences, shifted).
if (self.position > 0 and
(tokens1[self.start + self.position - 1] ==
tokens2[self.position - 1] + token2 or
tokens1[self.start + self.position - 1] ==
tokens2[self.position - 1] + '-' + token2 or
lemmas1[self.start + self.position - 1] ==
lemmas2[self.position - 1] + lem2 or
lemmas1[self.start + self.position - 1] ==
lemmas2[self.position - 1] + '-' + lem2)):
return False
if (self.position < len(tokens2) - 1 and
(tokens1[self.start + self.position + 1] ==
token2 + tokens2[self.position + 1] or
tokens1[self.start + self.position + 1] ==
token2 + '-' + tokens2[self.position + 1] or
lemmas1[self.start + self.position + 1] ==
lem2 + lemmas2[self.position + 1] or
lemmas1[self.start + self.position + 1] ==
lem2 + '-' + lemmas2[self.position + 1])):
return False
return True
