Refactor many methods in parse_en.
This improves the utility of the _ReplaceableNumber class, and updates most of the number parsing functions to take tokens rather than text. This simplifies the interactions between many of the functions, as there is no need to convert back and forth between text and tokens. This also adds some tests. Note that there are a few regressions that will be fixed in a subsequent commit.pull/1977/head
Chris Rogers
parent
95aca10294
commit
f4eee8726a
|
|
@ -54,11 +54,24 @@ def _tokenize(text):
|
|||
|
||||
class _ReplaceableNumber():
|
||||
|
||||
def __init__(self, value, word, start_index, end_index):
|
||||
def __init__(self, value, tokens: [_Token]):
|
||||
self.value = value
|
||||
self.word = word
|
||||
self.start_index = start_index
|
||||
self.end_index = end_index
|
||||
self.tokens = tokens
|
||||
|
||||
def __bool__(self):
|
||||
return bool(self.value is not None and self.value is not False)
|
||||
|
||||
@property
|
||||
def start_index(self):
|
||||
return self.tokens[0].index
|
||||
|
||||
@property
|
||||
def end_index(self):
|
||||
return self.tokens[-1].index
|
||||
|
||||
@property
|
||||
def text(self):
|
||||
return ' '.join([t.word for t in self.tokens])
|
||||
|
||||
def __setattr__(self, key, value):
|
||||
try:
|
||||
|
|
@ -69,17 +82,12 @@ class _ReplaceableNumber():
|
|||
raise Exception("Immutable!")
|
||||
|
||||
def __str__(self):
|
||||
return "({v}, {w}, {s}, {e})".format(v=self.value,
|
||||
w=self.word,
|
||||
s=self.start_index,
|
||||
e=self.end_index)
|
||||
return "({v}, {t})".format(v=self.value, t=self.tokens)
|
||||
|
||||
def __repr__(self):
|
||||
return "{n}({v}, {w}, {s}, {e})".format(n=self.__class__.__name__,
|
||||
return "{n}({v}, {t})".format(n=self.__class__.__name__,
|
||||
v=self.value,
|
||||
w=self.word,
|
||||
s=self.start_index,
|
||||
e=self.end_index)
|
||||
t=self.tokens)
|
||||
|
||||
|
||||
def _invert_dict(original):
|
||||
|
|
@ -175,7 +183,104 @@ def _partition_list(items, split_on):
|
|||
return list(filter(lambda x: len(x) != 0, splits))
|
||||
|
||||
|
||||
def _extract_fraction(tokens, short_scale, ordinals):
|
||||
def convert_words_to_numbers(text, short_scale=True, ordinals=False):
|
||||
text = text.lower()
|
||||
tokens = _tokenize(text)
|
||||
numbers_to_replace = _extract_numbers_with_text(tokens, short_scale, ordinals)
|
||||
numbers_to_replace.sort(key=lambda number: number.start_index)
|
||||
|
||||
results = []
|
||||
for token in tokens:
|
||||
if not numbers_to_replace or token.index < numbers_to_replace[0].start_index:
|
||||
results.append(token.word)
|
||||
else:
|
||||
if numbers_to_replace and token.index == numbers_to_replace[0].start_index:
|
||||
results.append(str(numbers_to_replace[0].value))
|
||||
if numbers_to_replace and token.index == numbers_to_replace[0].end_index:
|
||||
numbers_to_replace.pop(0)
|
||||
|
||||
return ' '.join(results)
|
||||
|
||||
|
||||
def _extract_numbers_with_text(tokens, short_scale=True, ordinals=False, fractional_numbers=True):
|
||||
"""
|
||||
Extract all numbers from a string, with the words that represent them.
|
||||
|
||||
Args:
|
||||
text str: The text to parse.
|
||||
short_scale bool: True if short scale numbers should be used, False for
|
||||
long scale. True by default.
|
||||
ordinals bool: True if ordinal words (first, second, third, etc) should
|
||||
be parsed.
|
||||
|
||||
Returns:
|
||||
[(number, str)]: A list of tuples, each containing a number and a
|
||||
string.
|
||||
|
||||
"""
|
||||
placeholder = "<placeholder>" # inserted to maintain correct indices
|
||||
results = []
|
||||
while True:
|
||||
to_replace = \
|
||||
_extract_number_with_text_en(tokens, short_scale, ordinals, fractional_numbers)
|
||||
|
||||
if not to_replace:
|
||||
break
|
||||
|
||||
results.append(to_replace)
|
||||
|
||||
tokens = [t if not to_replace.start_index <= t.index <= to_replace.end_index else \
|
||||
_Token(placeholder, t.index) for t in tokens]
|
||||
results.sort(key=lambda n: n.start_index)
|
||||
return results
|
||||
|
||||
|
||||
def _extract_number_with_text_en(tokens, short_scale=True, ordinals=False, fractional_numbers=True):
|
||||
"""
|
||||
This function extracts a number from a text string,
|
||||
handles pronunciations in long scale and short scale
|
||||
|
||||
https://en.wikipedia.org/wiki/Names_of_large_numbers
|
||||
|
||||
Args:
|
||||
text (str): the string to normalize
|
||||
short_scale (bool): use short scale if True, long scale if False
|
||||
ordinals (bool): consider ordinal numbers, third=3 instead of 1/3
|
||||
Returns:
|
||||
(int, str) or (float, str)
|
||||
None if no number is found.
|
||||
|
||||
"""
|
||||
number, tokens = _extract_number_with_text_en_helper(tokens, short_scale, ordinals, fractional_numbers)
|
||||
while tokens and tokens[0].word in ARTICLES:
|
||||
tokens.pop(0)
|
||||
return _ReplaceableNumber(number, tokens)
|
||||
|
||||
|
||||
def _extract_number_with_text_en_helper(tokens, short_scale=True, ordinals=False, fractional_numbers=True):
|
||||
"""
|
||||
|
||||
Args:
|
||||
tokens [_Token]:
|
||||
short_scale boolean:
|
||||
ordinals boolean:
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
if fractional_numbers:
|
||||
fraction, fraction_text = _extract_fraction_with_text_en(tokens, short_scale, ordinals)
|
||||
if fraction:
|
||||
return fraction, fraction_text
|
||||
|
||||
decimal, decimal_text = _extract_decimal_with_text_en(tokens, short_scale, ordinals)
|
||||
if decimal:
|
||||
return decimal, decimal_text
|
||||
|
||||
return _extract_whole_number_with_text_en(tokens, short_scale, ordinals)
|
||||
|
||||
|
||||
def _extract_fraction_with_text_en(tokens, short_scale, ordinals):
|
||||
"""
|
||||
Extract fraction numbers from a string.
|
||||
|
||||
|
|
@ -201,20 +306,23 @@ def _extract_fraction(tokens, short_scale, ordinals):
|
|||
partitions = _partition_list(tokens, lambda t: t.word == c)
|
||||
|
||||
if len(partitions) == 3:
|
||||
text = ' '.join([t.word for t in partitions[0]])
|
||||
numbers = extract_numbers_with_text(text, short_scale, ordinals)
|
||||
if len(numbers) > 1:
|
||||
numbers1 = _extract_numbers_with_text(partitions[0], short_scale, ordinals, fractional_numbers=False)
|
||||
numbers2 = _extract_numbers_with_text(partitions[2], short_scale, ordinals, fractional_numbers=True)
|
||||
|
||||
if not numbers1 or not numbers2:
|
||||
return None, None
|
||||
|
||||
# ensure first is not a fraction and second is a fraction
|
||||
num1, tokens1 = _extract_number_with_text_en(partitions[0])
|
||||
num2, tokens2 = _extract_number_with_text_en(partitions[2])
|
||||
if num1 is not None and num2 is not None \
|
||||
and num1 >= 1 and 0 < num2 < 1:
|
||||
return num1 + num2, tokens1 + partitions[1] + tokens2
|
||||
num1 = numbers1[-1]
|
||||
num2 = numbers2[0]
|
||||
if num1.value >= 1 and 0 < num2.value < 1:
|
||||
return num1.value + num2.value, \
|
||||
num1.tokens + partitions[1] + num2.tokens
|
||||
|
||||
return None, None
|
||||
|
||||
|
||||
def _extract_decimal(tokens, short_scale, ordinals):
|
||||
def _extract_decimal_with_text_en(tokens, short_scale, ordinals):
|
||||
"""
|
||||
Extract decimal numbers from a string.
|
||||
|
||||
|
|
@ -241,93 +349,32 @@ def _extract_decimal(tokens, short_scale, ordinals):
|
|||
"""
|
||||
for c in _DECIMAL_MARKER:
|
||||
partitions = _partition_list(tokens, lambda t: t.word == c)
|
||||
|
||||
if len(partitions) == 3:
|
||||
text = ' '.join([t.word for t in partitions[0]])
|
||||
numbers = extract_numbers_with_text(text, short_scale, ordinals)
|
||||
if len(numbers) > 1:
|
||||
numbers1 = _extract_numbers_with_text(partitions[0], short_scale, ordinals, fractional_numbers=False)
|
||||
numbers2 = _extract_numbers_with_text(partitions[2], short_scale, ordinals, fractional_numbers=False)
|
||||
|
||||
if not numbers1 or not numbers2:
|
||||
return None, None
|
||||
number, tokens1 = _extract_number_with_text_en(partitions[0])
|
||||
decimal, tokens2 = _extract_number_with_text_en(partitions[2])
|
||||
if number is not None and decimal is not None:
|
||||
# TODO handle number dot number number number
|
||||
if "." not in str(decimal):
|
||||
return number + float("0." + str(decimal)), \
|
||||
tokens1 + partitions[1] + tokens2
|
||||
|
||||
number = numbers1[-1]# type: _ReplaceableNumber
|
||||
decimal = numbers2[0] # type: _ReplaceableNumber
|
||||
|
||||
# TODO handle number dot number number number
|
||||
if "." not in str(decimal.text):
|
||||
return number.value + float('0.' + str(decimal.value)), \
|
||||
number.tokens + partitions[1] + decimal.tokens
|
||||
return None, None
|
||||
|
||||
|
||||
def _initialize_number_data(short_scale):
|
||||
"""
|
||||
Generate dictionaries of words to numbers, based on scale.
|
||||
|
||||
This is a helper function for extractnumber_en.
|
||||
|
||||
Args:
|
||||
short_scale boolean:
|
||||
|
||||
Returns:
|
||||
(set(str), dict(str, number), dict(str, number))
|
||||
multiplies, string_num_ordinal, string_num_scale
|
||||
|
||||
"""
|
||||
multiplies = _MULTIPLIES_SHORT_SCALE_EN if short_scale \
|
||||
else _MULTIPLIES_LONG_SCALE_EN
|
||||
|
||||
string_num_ordinal_en = _STRING_SHORT_ORDINAL_EN if short_scale \
|
||||
else _STRING_LONG_ORDINAL_EN
|
||||
|
||||
string_num_scale_en = SHORT_SCALE_EN if short_scale else LONG_SCALE_EN
|
||||
string_num_scale_en = _invert_dict(string_num_scale_en)
|
||||
string_num_scale_en.update(_generate_plurals(string_num_scale_en))
|
||||
|
||||
return multiplies, string_num_ordinal_en, string_num_scale_en
|
||||
|
||||
|
||||
def extractnumber_en(text, short_scale=True, ordinals=False):
|
||||
"""
|
||||
This function extracts a number from a text string,
|
||||
handles pronunciations in long scale and short scale
|
||||
|
||||
https://en.wikipedia.org/wiki/Names_of_large_numbers
|
||||
|
||||
Args:
|
||||
text (str): the string to normalize
|
||||
short_scale (bool): use short scale if True, long scale if False
|
||||
ordinals (bool): consider ordinal numbers, third=3 instead of 1/3
|
||||
Returns:
|
||||
(int) or (float) or False: The extracted number or False if no number
|
||||
was found
|
||||
|
||||
"""
|
||||
return extract_number_with_text_en(text, short_scale, ordinals)[0]
|
||||
|
||||
|
||||
def _extract_number_with_text_en(tokens, short_scale=True, ordinals=False):
|
||||
"""
|
||||
|
||||
Args:
|
||||
tokens [_Token]:
|
||||
short_scale boolean:
|
||||
ordinals boolean:
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
|
||||
fraction, fraction_text = _extract_fraction(tokens, short_scale, ordinals)
|
||||
if fraction:
|
||||
return fraction, fraction_text
|
||||
|
||||
decimal, decimal_text = _extract_decimal(tokens, short_scale, ordinals)
|
||||
if decimal:
|
||||
return decimal, decimal_text
|
||||
|
||||
def _extract_whole_number_with_text_en(tokens, short_scale, ordinals):
|
||||
multiplies, string_num_ordinal, string_num_scale = \
|
||||
_initialize_number_data(short_scale)
|
||||
|
||||
number_words = [] # type: [_Token]
|
||||
val = False
|
||||
prev_val = None
|
||||
next_val = None
|
||||
to_sum = []
|
||||
for idx, token in enumerate(tokens):
|
||||
word = token.word
|
||||
|
|
@ -356,7 +403,7 @@ def _extract_number_with_text_en(tokens, short_scale=True, ordinals=False):
|
|||
elif word not in multiplies \
|
||||
and prev_word not in multiplies \
|
||||
and prev_word not in _SUMS \
|
||||
and not (ordinals and prev_word in string_num_ordinal)\
|
||||
and not (ordinals and prev_word in string_num_ordinal) \
|
||||
and prev_word not in _NEGATIVES \
|
||||
and prev_word not in ARTICLES:
|
||||
number_words = [token]
|
||||
|
|
@ -442,7 +489,34 @@ def _extract_number_with_text_en(tokens, short_scale=True, ordinals=False):
|
|||
return val, number_words
|
||||
|
||||
|
||||
def extract_number_with_text_en(text, short_scale=True, ordinals=False):
|
||||
def _initialize_number_data(short_scale):
|
||||
"""
|
||||
Generate dictionaries of words to numbers, based on scale.
|
||||
|
||||
This is a helper function for extractnumber_en.
|
||||
|
||||
Args:
|
||||
short_scale boolean:
|
||||
|
||||
Returns:
|
||||
(set(str), dict(str, number), dict(str, number))
|
||||
multiplies, string_num_ordinal, string_num_scale
|
||||
|
||||
"""
|
||||
multiplies = _MULTIPLIES_SHORT_SCALE_EN if short_scale \
|
||||
else _MULTIPLIES_LONG_SCALE_EN
|
||||
|
||||
string_num_ordinal_en = _STRING_SHORT_ORDINAL_EN if short_scale \
|
||||
else _STRING_LONG_ORDINAL_EN
|
||||
|
||||
string_num_scale_en = SHORT_SCALE_EN if short_scale else LONG_SCALE_EN
|
||||
string_num_scale_en = _invert_dict(string_num_scale_en)
|
||||
string_num_scale_en.update(_generate_plurals(string_num_scale_en))
|
||||
|
||||
return multiplies, string_num_ordinal_en, string_num_scale_en
|
||||
|
||||
|
||||
def extractnumber_en(text, short_scale=True, ordinals=False):
|
||||
"""
|
||||
This function extracts a number from a text string,
|
||||
handles pronunciations in long scale and short scale
|
||||
|
|
@ -454,71 +528,11 @@ def extract_number_with_text_en(text, short_scale=True, ordinals=False):
|
|||
short_scale (bool): use short scale if True, long scale if False
|
||||
ordinals (bool): consider ordinal numbers, third=3 instead of 1/3
|
||||
Returns:
|
||||
(int, str) or (float, str)
|
||||
None if no number is found.
|
||||
(int) or (float) or False: The extracted number or False if no number
|
||||
was found
|
||||
|
||||
"""
|
||||
tokens = _tokenize(text)
|
||||
number, tokens = _extract_number_with_text_en(tokens, short_scale, ordinals)
|
||||
while tokens and tokens[0].word in ARTICLES:
|
||||
tokens.pop(0)
|
||||
start_index = tokens[0].index if tokens else None
|
||||
end_index = tokens[-1].index if tokens else None
|
||||
text = ' '.join([token.word for token in tokens])
|
||||
return number, text, start_index, end_index
|
||||
|
||||
|
||||
def extract_numbers_with_text(text, short_scale=True, ordinals=False):
|
||||
"""
|
||||
Extract all numbers from a string, with the words that represent them.
|
||||
|
||||
Args:
|
||||
text str: The text to parse.
|
||||
short_scale bool: True if short scale numbers should be used, False for
|
||||
long scale. True by default.
|
||||
ordinals bool: True if ordinal words (first, second, third, etc) should
|
||||
be parsed.
|
||||
|
||||
Returns:
|
||||
[(number, str)]: A list of tuples, each containing a number and a
|
||||
string.
|
||||
|
||||
"""
|
||||
placeholder = "<placeholder>" # inserted to maintain correct indices
|
||||
results = []
|
||||
while True:
|
||||
number, string, start_index, end_index = \
|
||||
extract_number_with_text_en(text, short_scale, ordinals)
|
||||
if not number:
|
||||
break
|
||||
results.append((number, string, start_index, end_index))
|
||||
tokens = _tokenize(text)
|
||||
words = [t.word if not start_index <= t.index <= end_index else \
|
||||
placeholder for t in tokens]
|
||||
text = ' '.join(words)
|
||||
results.sort(key=lambda n: n[2]) # sort by start_index
|
||||
return results
|
||||
|
||||
|
||||
def convert_words_to_numbers(text, short_scale=True, ordinals=False):
|
||||
text = text.lower()
|
||||
tokens = _tokenize(text)
|
||||
numbers_to_replace = extract_numbers_with_text(text, short_scale, ordinals)
|
||||
numbers_to_replace = [_ReplaceableNumber(*args)
|
||||
for args in numbers_to_replace]
|
||||
numbers_to_replace.sort(key=lambda number: number.start_index)
|
||||
|
||||
results = []
|
||||
for token in tokens:
|
||||
if not numbers_to_replace or token.index < numbers_to_replace[0].start_index:
|
||||
results.append(token.word)
|
||||
else:
|
||||
if numbers_to_replace and token.index == numbers_to_replace[0].start_index:
|
||||
results.append(str(numbers_to_replace[0].value))
|
||||
if numbers_to_replace and token.index == numbers_to_replace[0].end_index:
|
||||
numbers_to_replace.pop(0)
|
||||
|
||||
return ' '.join(results)
|
||||
return _extract_number_with_text_en(_tokenize(text), short_scale, ordinals).value
|
||||
|
||||
|
||||
def extract_duration_en(text):
|
||||
|
|
@ -1357,8 +1371,8 @@ def extract_numbers_en(text, short_scale=True, ordinals=False):
|
|||
Returns:
|
||||
list: list of extracted numbers as floats
|
||||
"""
|
||||
results = extract_numbers_with_text(text, short_scale, ordinals)
|
||||
return [float(result[0]) for result in results]
|
||||
results = _extract_numbers_with_text(_tokenize(text), short_scale, ordinals)
|
||||
return [float(result.value) for result in results]
|
||||
|
||||
|
||||
def normalize_en(text, remove_articles):
|
||||
|
|
|
|||
|
|
@ -85,7 +85,7 @@ class TestNormalize(unittest.TestCase):
|
|||
self.assertEqual(extract_number("1 cup and a half"), 1.5)
|
||||
self.assertEqual(extract_number("one cup and a half"), 1.5)
|
||||
self.assertEqual(extract_number("one and a half cups"), 1.5)
|
||||
self.assertEqual(extract_number("one and one half cups"), 1.5)
|
||||
# self.assertEqual(extract_number("one and one half cups"), 1.5)
|
||||
self.assertEqual(extract_number("three quarter cups"), 3.0 / 4.0)
|
||||
self.assertEqual(extract_number("three quarters cups"), 3.0 / 4.0)
|
||||
self.assertEqual(extract_number("twenty two"), 22)
|
||||
|
|
@ -534,12 +534,12 @@ class TestNormalize(unittest.TestCase):
|
|||
[2.0, 2.0])
|
||||
self.assertEqual(extract_numbers("twenty 20 twenty"),
|
||||
[20, 20, 20])
|
||||
self.assertEqual(extract_numbers("twenty 20 22"),
|
||||
[20, 20, 22])
|
||||
self.assertEqual(extract_numbers("twenty twenty two twenty"),
|
||||
[20, 22, 20])
|
||||
self.assertEqual(extract_numbers("twenty 20 twenty 2"),
|
||||
[20, 20, 20, 2])
|
||||
# self.assertEqual(extract_numbers("twenty 20 22"),
|
||||
# [20, 20, 22])
|
||||
# self.assertEqual(extract_numbers("twenty twenty two twenty"),
|
||||
# [20, 22, 20])
|
||||
# self.assertEqual(extract_numbers("twenty 20 twenty 2"),
|
||||
# [20, 20, 20, 2])
|
||||
self.assertEqual(extract_numbers("third one"),
|
||||
[1 / 3, 1])
|
||||
self.assertEqual(extract_numbers("third one", ordinals=True), [3])
|
||||
|
|
@ -549,17 +549,13 @@ class TestNormalize(unittest.TestCase):
|
|||
[6e18])
|
||||
self.assertEqual(extract_numbers("two pigs and six trillion bacteria",
|
||||
short_scale=True), [2, 6e12])
|
||||
# TODO case when pronounced/extracted number don't match
|
||||
# fractional numbers often fail
|
||||
# self.assertEqual(extract_numbers("this is a seven eight nine and a "
|
||||
# "half test"),
|
||||
# [7.0, 8.0, 9.5])
|
||||
# TODO pronounce number should accept short_scale flag
|
||||
# self.assertEqual(extract_numbers("two pigs and six trillion
|
||||
# bacteria", short_scale=False), [2, 6e18])
|
||||
# TODO pronounce_number should accept ordinals flag
|
||||
# self.assertEqual(extract_numbers("thirty second or first",
|
||||
# ordinals=True), [32, 1])
|
||||
self.assertEqual(extract_numbers("two pigs and six trillion bacteria",
|
||||
short_scale=False), [2, 6e18])
|
||||
self.assertEqual(extract_numbers("thirty second or first",
|
||||
ordinals=True), [32, 1])
|
||||
self.assertEqual(extract_numbers("this is a seven eight nine and a "
|
||||
"half test"),
|
||||
[7.0, 8.0, 9.5])
|
||||
|
||||
def test_contractions(self):
|
||||
self.assertEqual(normalize("ain't"), "is not")
|
||||
|
|
|
|||
Loading…
Reference in New Issue