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
"""
string_num_en = {
"half": 0.5,
"halves": 0.5,
"couple": 2,
"hundred": 100,
"hundreds": 100,
"thousand": 1000,
"thousands": 1000,
"million": 1000000,
'millions': 1000000}
string_num_ordinal_en = {}
for num in NUM_STRING_EN:
num_string = NUM_STRING_EN[num]
string_num_en[num_string] = num
# first, second...
if ordinals:
if short_scale:
for num in SHORT_ORDINAL_STRING_EN:
num_string = SHORT_ORDINAL_STRING_EN[num]
string_num_ordinal_en[num_string] = num
string_num_en[num_string] = num
else:
for num in LONG_ORDINAL_STRING_EN:
num_string = LONG_ORDINAL_STRING_EN[num]
string_num_ordinal_en[num_string] = num
string_num_en[num_string] = num
# negate next number (-2 = 0 - 2)
negatives = ["negative", "minus"]
# sum the next number (twenty two = 20 + 2)
sums = ['twenty', 'thirty', 'forty', 'fifty', 'sixty', 'seventy', 'eighty',
'ninety']
# multiply the previous number (one hundred = 1 * 100)
multiplies = ["hundred", "thousand", "hundreds", "thousands", "million",
"millions"]
# split sentence parse separately and sum ( 2 and a half = 2 + 0.5 )
fraction_marker = [" and "]
# decimal marker ( 1 point 5 = 1 + 0.5)
decimal_marker = [" point ", " dot "]
if short_scale:
for num in SHORT_SCALE_EN:
num_string = SHORT_SCALE_EN[num]
string_num_en[num_string] = num
string_num_en[num_string + "s"] = num
multiplies.append(num_string)
multiplies.append(num_string + "s")
else:
for num in LONG_SCALE_EN:
num_string = LONG_SCALE_EN[num]
string_num_en[num_string] = num
string_num_en[num_string + "s"] = num
multiplies.append(num_string)
multiplies.append(num_string + "s")
# 2 and 3/4
for c in fraction_marker:
components = text.split(c)
if len(components) == 2:
# ensure first is not a fraction and second is a fraction
num1 = extractnumber_en(components[0])
num2 = extractnumber_en(components[1])
if num1 is not None and num2 is not None \
and num1 >= 1 and 0 < num2 < 1:
return num1 + num2
# 2 point 5
for c in decimal_marker:
components = text.split(c)
if len(components) == 2:
number = extractnumber_en(components[0])
decimal = extractnumber_en(components[1])
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))
aWords = text.split()
aWords = [word for word in aWords if word not in ["the", "a", "an"]]
val = False
prev_val = None
to_sum = []
for idx, word in enumerate(aWords):
if not word:
continue
prev_word = aWords[idx - 1] if idx > 0 else ""
next_word = aWords[idx + 1] if idx + 1 < len(aWords) else ""
# is this word already a number ?
if is_numeric(word):
# if word.isdigit(): # doesn't work with decimals
val = float(word)
# is this word the name of a number ?
if word in string_num_en:
val = string_num_en[word]
# is the prev word an ordinal number and current word is one?
# second one, third one
if ordinals and prev_word in string_num_ordinal_en and val is 1:
val = prev_val
# is the prev word a number and should we sum it?
# twenty two, fifty six
if prev_word in sums and word in string_num_en:
if val and val < 10:
val = prev_val + val
# is the prev word a number and should we multiply it?
# twenty hundred, six hundred
if word in multiplies:
if not prev_val:
prev_val = 1
val = prev_val * val
# is this a spoken fraction?
# half cup
if val is False:
val = isFractional_en(word, short_scale=short_scale)
# 2 fifths
if not ordinals:
next_value = isFractional_en(next_word, short_scale=short_scale)
if next_value:
if not val:
val = 1
val = val * next_value
# is this a negative number?
if val and prev_word and prev_word in negatives:
val = 0 - val
# let's make sure it isn't a fraction
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
else:
prev_val = val
# handle long numbers
# six hundred sixty six
# two million five hundred thousand
if word in multiplies and next_word not in multiplies:
to_sum.append(val)
val = 0
prev_val = 0
if val is not None:
for v in to_sum:
val = val + v
return val
def extractnumber_it(text):
"""
Questa funzione prepara il testo dato per l'analisi rendendo
numeri testuali come interi o frazioni.
In italiano non è un modo abituale ma può essere interessante
per Mycroft
E' la versione portoghese riadattata in italiano
args:
text (str): la stringa da normalizzare
Ritorna:
(int) o (float): il valore del numero estratto
"""
aWords = text.split()
count = 0
result = None
while count < len(aWords):
val = 0
word = aWords[count]
next_next_word = None
if count + 1 < len(aWords):
next_word = aWords[count + 1]
if count + 2 < len(aWords):
next_next_word = aWords[count + 2]
else:
next_word = None
# is current word a number?
if word in it_numbers:
if word == "mila":
val = it_numbers[word]
val = result * val
result = 0
else:
val = it_numbers[word]
elif word.isdigit(): # doesn't work with decimals
val = int(word)
elif is_numeric(word):
val = float(word)
elif isFractional_it(word):
if not result:
result = 1
result = result * isFractional_it(word)
# "un terzo" is 1/3 but "il terzo" is 3
if aWords[count - 1] == "il":
result = 1.0 // isFractional_it(word)
count += 1
continue
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
if not val:
# cerca numero composto come ventuno ventitre centoventi"
val = extractnumber_long_it(word)
if val:
if result is None:
result = 0
# handle fractions
# if next_word != "avos":
result += val
# else:
# result = float(result) / float(val)
if next_word is None:
break
# number word and fraction
ands = ["e"]
if next_word in ands:
zeros = 0
if result is None:
count += 1
continue
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_it(newText[:-1])
if afterAndVal:
if result < afterAndVal or result < 20:
while afterAndVal > 1:
afterAndVal = afterAndVal / 10.0
for word in newWords:
if word == "zero" or word == "0":
zeros += 1
else:
break
for _ in range(0, zeros):
afterAndVal = afterAndVal / 10.0
result += afterAndVal
break
elif next_next_word is not None:
if next_next_word in ands:
newWords = aWords[count + 3:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_it(newText[:-1])
if afterAndVal:
if result is None:
result = 0
result += afterAndVal
break
decimals = ["punto", "virgola", ".", ","]
if next_word in decimals:
zeros = 0
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
for word in newWords:
if word == "zero" or word == "0":
zeros += 1
else:
break
afterDotVal = str(extractnumber_it(newText[:-1]))
afterDotVal = zeros * "0" + afterDotVal
result = float(str(result) + "." + afterDotVal)
break
count += 1
if result is None:
return False
# Return the $str with the number related words removed
# (now empty strings, so strlen == 0)
# aWords = [word for word in aWords if len(word) > 0]
# text = ' '.join(aWords)
if "." in str(result):
integer, dec = str(result).split(".")
# cast float to int
if dec == "0":
result = int(integer)
return result
def extractnumber_sv(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
"""
aWords = text.split()
and_pass = False
valPreAnd = False
val = False
count = 0
while count < len(aWords):
word = aWords[count]
if is_numeric(word):
val = float(word)
elif word == "första":
val = 1
elif word == "andra":
val = 2
elif word == "tredje":
val = 3
elif word == "fjärde":
val = 4
elif word == "femte":
val = 5
elif word == "sjätte":
val = 6
elif is_fractional_sv(word):
val = is_fractional_sv(word)
else:
if word == "en":
val = 1
if word == "ett":
val = 1
elif word == "två":
val = 2
elif word == "tre":
val = 3
elif word == "fyra":
val = 4
elif word == "fem":
val = 5
elif word == "sex":
val = 6
elif word == "sju":
val = 7
elif word == "åtta":
val = 8
elif word == "nio":
val = 9
elif word == "tio":
val = 10
if val:
if count < (len(aWords) - 1):
wordNext = aWords[count + 1]
else:
wordNext = ""
valNext = is_fractional_sv(wordNext)
if valNext:
val = val * valNext
aWords[count + 1] = ""
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
elif and_pass:
# added to value, quit here
val = valPreAnd
break
else:
count += 1
continue
aWords[count] = ""
if and_pass:
aWords[count - 1] = '' # remove "och"
val += valPreAnd
elif count + 1 < len(aWords) and aWords[count + 1] == 'och':
and_pass = True
valPreAnd = val
val = False
count += 2
continue
elif count + 2 < len(aWords) and aWords[count + 2] == 'och':
and_pass = True
valPreAnd = val
val = False
count += 3
continue
break
if not val:
return False
return val
def extractnumber_fr(text):
"""Takes in a string and extracts a number.
Args:
text (str): the string to extract a number from
Returns:
(str): The number extracted or the original text.
"""
# normalize text, keep articles for ordinals versus fractionals
text = normalize_fr(text, False)
# split words by whitespace
aWords = text.split()
count = 0
result = None
add = False
while count < len(aWords):
val = None
word = aWords[count]
wordNext = ""
wordPrev = ""
if count < (len(aWords) - 1):
wordNext = aWords[count + 1]
if count > 0:
wordPrev = aWords[count - 1]
if word in articles_fr:
count += 1
continue
if word in ["et", "plus", "+"]:
count += 1
add = True
continue
# is current word a numeric number?
if word.isdigit():
val = int(word)
count += 1
elif is_numeric(word):
val = float(word)
count += 1
elif wordPrev in articles_fr and getOrdinal_fr(word):
val = getOrdinal_fr(word)
count += 1
# is current word the denominator of a fraction?
elif isFractional_fr(word):
val = isFractional_fr(word)
count += 1
# is current word the numerator of a fraction?
if val and wordNext:
valNext = isFractional_fr(wordNext)
if valNext:
val = float(val) * valNext
count += 1
if not val:
count += 1
# is current word a numeric fraction like "2/3"?
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
# is current word followed by a decimal value?
if wordNext == "virgule":
zeros = 0
newWords = aWords[count + 1:]
# count the number of zeros after the decimal sign
for word in newWords:
if word == "zéro" or word == "0":
zeros += 1
else:
break
afterDotVal = None
# extract the number after the zeros
if newWords[zeros].isdigit():
afterDotVal = newWords[zeros]
countDot = count + zeros + 2
# if a number was extracted (since comma is also a
# punctuation sign)
if afterDotVal:
count = countDot
if not val:
val = 0
# add the zeros
afterDotString = zeros * "0" + afterDotVal
val = float(str(val) + "." + afterDotString)
if val:
if add:
result += val
add = False
else:
result = val
# if result == False:
if not result:
return normalize_fr(text, True)
return result
def extractnumber_da(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
undefined articles cannot be suppressed in German:
'ein Pferd' means 'one horse' and 'a horse'
"""
aWords = text.split()
aWords = [word for word in aWords if word not in ["den", "det"]]
and_pass = False
valPreAnd = False
val = False
count = 0
while count < len(aWords):
word = aWords[count]
if is_numeric(word):
if word.isdigit(): # doesn't work with decimals
val = float(word)
elif isFractional_da(word):
val = isFractional_da(word)
elif isOrdinal_da(word):
val = isOrdinal_da(word)
else:
if word in da_numbers:
val = da_numbers[word]
if count < (len(aWords) - 1):
wordNext = aWords[count + 1]
else:
wordNext = ""
valNext = isFractional_da(wordNext)
if valNext:
val = val * valNext
aWords[count + 1] = ""
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
elif and_pass:
# added to value, quit here
val = valPreAnd
break
else:
count += 1
continue
aWords[count] = ""
if and_pass:
aWords[count - 1] = '' # remove "og"
val += valPreAnd
elif count + 1 < len(aWords) and aWords[count + 1] == 'og':
and_pass = True
valPreAnd = val
val = False
count += 2
continue
elif count + 2 < len(aWords) and aWords[count + 2] == 'og':
and_pass = True
valPreAnd = val
val = False
count += 3
continue
break
if not val:
return False
return val
def extractnumber_da(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
undefined articles cannot be suppressed in German:
'ein Pferd' means 'one horse' and 'a horse'
"""
aWords = text.split()
aWords = [word for word in aWords if
word not in ["den", "det"]]
and_pass = False
valPreAnd = False
val = False
count = 0
while count < len(aWords):
word = aWords[count]
if is_numeric(word):
if word.isdigit(): # doesn't work with decimals
val = float(word)
elif isFractional_da(word):
val = isFractional_da(word)
elif isOrdinal_da(word):
val = isOrdinal_da(word)
else:
if word in da_numbers:
val = da_numbers[word]
if count < (len(aWords) - 1):
wordNext = aWords[count + 1]
else:
wordNext = ""
valNext = isFractional_da(wordNext)
if valNext:
val = val * valNext
aWords[count + 1] = ""
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
elif and_pass:
# added to value, quit here
val = valPreAnd
break
else:
count += 1
continue
aWords[count] = ""
if and_pass:
aWords[count - 1] = '' # remove "og"
val += valPreAnd
elif count + 1 < len(aWords) and aWords[count + 1] == 'og':
and_pass = True
valPreAnd = val
val = False
count += 2
continue
elif count + 2 < len(aWords) and aWords[count + 2] == 'og':
and_pass = True
valPreAnd = val
val = False
count += 3
continue
break
if not val:
return False
return val
def extractnumber_it(text):
"""
Questa funzione prepara il testo dato per l'analisi rendendo
numeri testuali come interi o frazioni.
In italiano non è un modo abituale ma può essere interessante
per Mycroft
E' la versione portoghese riadattata in italiano
args:
text (str): la stringa da normalizzare
Ritorna:
(int) o (float): il valore del numero estratto
"""
aWords = text.split()
count = 0
result = None
while count < len(aWords):
val = 0
word = aWords[count]
next_next_word = None
if count + 1 < len(aWords):
next_word = aWords[count + 1]
if count + 2 < len(aWords):
next_next_word = aWords[count + 2]
else:
next_word = None
# is current word a number?
if word in it_numbers:
if word == "mila":
val = it_numbers[word]
val = result * val
result = 0
else:
val = it_numbers[word]
elif word.isdigit(): # doesn't work with decimals
val = int(word)
elif is_numeric(word):
val = float(word)
elif isFractional_it(word):
if not result:
result = 1
result = result * isFractional_it(word)
# "un terzo" is 1/3 but "il terzo" is 3
if aWords[count - 1] == "il":
result = 1.0 // isFractional_it(word)
count += 1
continue
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
if not val:
# cerca numero composto come ventuno ventitre centoventi"
val = extractnumber_long_it(word)
if val:
if result is None:
result = 0
# handle fractions
# if next_word != "avos":
result += val
# else:
# result = float(result) / float(val)
if next_word is None:
break
# number word and fraction
ands = ["e"]
if next_word in ands:
zeros = 0
if result is None:
count += 1
continue
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_it(newText[:-1])
if afterAndVal:
if result < afterAndVal or result < 20:
while afterAndVal > 1:
afterAndVal = afterAndVal / 10.0
for word in newWords:
if word == "zero" or word == "0":
zeros += 1
else:
break
for _ in range(0, zeros):
afterAndVal = afterAndVal / 10.0
result += afterAndVal
break
elif next_next_word is not None:
if next_next_word in ands:
newWords = aWords[count + 3:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_it(newText[:-1])
if afterAndVal:
if result is None:
result = 0
result += afterAndVal
break
decimals = ["punto", "virgola", ".", ","]
if next_word in decimals:
zeros = 0
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
for word in newWords:
if word == "zero" or word == "0":
zeros += 1
else:
break
afterDotVal = str(extractnumber_it(newText[:-1]))
afterDotVal = zeros * "0" + afterDotVal
result = float(str(result) + "." + afterDotVal)
break
count += 1
if result is None:
return False
# Return the $str with the number related words removed
# (now empty strings, so strlen == 0)
# aWords = [word for word in aWords if len(word) > 0]
# text = ' '.join(aWords)
if "." in str(result):
integer, dec = str(result).split(".")
# cast float to int
if dec == "0":
result = int(integer)
return result
def extractnumber_pt(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
"""
aWords = text.split()
count = 0
result = None
while count < len(aWords):
val = 0
word = aWords[count]
next_next_word = None
if count + 1 < len(aWords):
next_word = aWords[count + 1]
if count + 2 < len(aWords):
next_next_word = aWords[count + 2]
else:
next_word = None
# is current word a number?
if word in pt_numbers:
val = pt_numbers[word]
elif word.isdigit(): # doesn't work with decimals
val = int(word)
elif is_numeric(word):
val = float(word)
elif isFractional_pt(word):
if not result:
result = 1
result = result * isFractional_pt(word)
count += 1
continue
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
if val:
if result is None:
result = 0
# handle fractions
if next_word != "avos":
result += val
else:
result = float(result) / float(val)
if next_word is None:
break
# number word and fraction
ands = ["e"]
if next_word in ands:
zeros = 0
if result is None:
count += 1
continue
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_pt(newText[:-1])
if afterAndVal:
if result < afterAndVal or result < 20:
while afterAndVal > 1:
afterAndVal = afterAndVal / 10.0
for word in newWords:
if word == "zero" or word == "0":
zeros += 1
else:
break
for _ in range(0, zeros):
afterAndVal = afterAndVal / 10.0
result += afterAndVal
break
elif next_next_word is not None:
if next_next_word in ands:
newWords = aWords[count + 3:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_pt(newText[:-1])
if afterAndVal:
if result is None:
result = 0
result += afterAndVal
break
decimals = ["ponto", "virgula", u"v�rgula", ".", ","]
if next_word in decimals:
zeros = 0
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
for word in newWords:
if word == "zero" or word == "0":
zeros += 1
else:
break
afterDotVal = str(extractnumber_pt(newText[:-1]))
afterDotVal = zeros * "0" + afterDotVal
result = float(str(result) + "." + afterDotVal)
break
count += 1
if result is None:
return False
# Return the $str with the number related words removed
# (now empty strings, so strlen == 0)
# aWords = [word for word in aWords if len(word) > 0]
# text = ' '.join(aWords)
if "." in str(result):
integer, dec = str(result).split(".")
# cast float to int
if dec == "0":
result = int(integer)
return result
def extractnumber_es(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
"""
aWords = text.split()
count = 0
result = None
while count < len(aWords):
val = 0
word = aWords[count]
next_next_word = None
if count + 1 < len(aWords):
next_word = aWords[count + 1]
if count + 2 < len(aWords):
next_next_word = aWords[count + 2]
else:
next_word = None
# is current word a number?
if word in es_numbers:
val = es_numbers[word]
elif word.isdigit(): # doesn't work with decimals
val = int(word)
elif is_numeric(word):
val = float(word)
elif isFractional_es(word):
if not result:
result = 1
result = result * isFractional_es(word)
count += 1
continue
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
if val:
if result is None:
result = 0
# handle fractions
if next_word != "avos":
result += val
else:
result = float(result) / float(val)
if next_word is None:
break
# number word and fraction
ands = ["e"]
if next_word in ands:
zeros = 0
if result is None:
count += 1
continue
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_es(newText[:-1])
if afterAndVal:
if result < afterAndVal or result < 20:
while afterAndVal > 1:
afterAndVal = afterAndVal / 10.0
for word in newWords:
if word == "cero" or word == "0":
zeros += 1
else:
break
for _ in range(0, zeros):
afterAndVal = afterAndVal / 10.0
result += afterAndVal
break
elif next_next_word is not None:
if next_next_word in ands:
newWords = aWords[count + 3:]
newText = ""
for word in newWords:
newText += word + " "
afterAndVal = extractnumber_es(newText[:-1])
if afterAndVal:
if result is None:
result = 0
result += afterAndVal
break
decimals = ["punto", "coma", ".", ","]
if next_word in decimals:
zeros = 0
newWords = aWords[count + 2:]
newText = ""
for word in newWords:
newText += word + " "
for word in newWords:
if word == "cero" or word == "0":
zeros += 1
else:
break
afterDotVal = str(extractnumber_es(newText[:-1]))
afterDotVal = zeros * "0" + afterDotVal
result = float(str(result) + "." + afterDotVal)
break
count += 1
if result is None:
return False
# Return the $str with the number related words removed
# (now empty strings, so strlen == 0)
# aWords = [word for word in aWords if len(word) > 0]
# text = ' '.join(aWords)
if "." in str(result):
integer, dec = str(result).split(".")
# cast float to int
if dec == "0":
result = int(integer)
return result
def extractnumber_en(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
"""
aWords = text.split()
aWords = [word for word in aWords if word not in ["the", "a", "an"]]
and_pass = False
valPreAnd = False
val = False
count = 0
while count < len(aWords):
word = aWords[count]
if is_numeric(word):
# if word.isdigit(): # doesn't work with decimals
val = float(word)
elif word == "first":
val = 1
elif word == "second":
val = 2
elif isFractional_en(word):
val = isFractional_en(word)
else:
if word == "one":
val = 1
elif word == "two":
val = 2
elif word == "three":
val = 3
elif word == "four":
val = 4
elif word == "five":
val = 5
elif word == "six":
val = 6
elif word == "seven":
val = 7
elif word == "eight":
val = 8
elif word == "nine":
val = 9
elif word == "ten":
val = 10
if val:
if count < (len(aWords) - 1):
wordNext = aWords[count + 1]
else:
wordNext = ""
valNext = isFractional_en(wordNext)
if valNext:
val = val * valNext
aWords[count + 1] = ""
# if val == False:
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
elif and_pass:
# added to value, quit here
val = valPreAnd
break
else:
count += 1
continue
aWords[count] = ""
if and_pass:
aWords[count - 1] = '' # remove "and"
val += valPreAnd
elif count + 1 < len(aWords) and aWords[count + 1] == 'and':
and_pass = True
valPreAnd = val
val = False
count += 2
continue
elif count + 2 < len(aWords) and aWords[count + 2] == 'and':
and_pass = True
valPreAnd = val
val = False
count += 3
continue
break
# if val == False:
if not val:
return False
# Return the string with the number related words removed
# (now empty strings, so strlen == 0)
aWords = [word for word in aWords if len(word) > 0]
text = ' '.join(aWords)
return val
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
"""
string_num_en = {
"half": 0.5,
"halves": 0.5,
"couple": 2,
"hundred": 100,
"hundreds": 100,
"thousand": 1000,
"thousands": 1000,
"million": 1000000,
'millions': 1000000
}
string_num_ordinal_en = {}
for num in NUM_STRING_EN:
num_string = NUM_STRING_EN[num]
string_num_en[num_string] = num
# first, second...
if ordinals:
if short_scale:
for num in SHORT_ORDINAL_STRING_EN:
num_string = SHORT_ORDINAL_STRING_EN[num]
string_num_ordinal_en[num_string] = num
string_num_en[num_string] = num
else:
for num in LONG_ORDINAL_STRING_EN:
num_string = LONG_ORDINAL_STRING_EN[num]
string_num_ordinal_en[num_string] = num
string_num_en[num_string] = num
# negate next number (-2 = 0 - 2)
negatives = ["negative", "minus"]
# sum the next number (twenty two = 20 + 2)
sums = [
'twenty', 'thirty', 'forty', 'fifty', 'sixty', 'seventy', 'eighty',
'ninety'
]
# multiply the previous number (one hundred = 1 * 100)
multiplies = [
"hundred", "thousand", "hundreds", "thousands", "million", "millions"
]
# split sentence parse separately and sum ( 2 and a half = 2 + 0.5 )
fraction_marker = [" and "]
# decimal marker ( 1 point 5 = 1 + 0.5)
decimal_marker = [" point ", " dot "]
if short_scale:
for num in SHORT_SCALE_EN:
num_string = SHORT_SCALE_EN[num]
string_num_en[num_string] = num
string_num_en[num_string + "s"] = num
multiplies.append(num_string)
multiplies.append(num_string + "s")
else:
for num in LONG_SCALE_EN:
num_string = LONG_SCALE_EN[num]
string_num_en[num_string] = num
string_num_en[num_string + "s"] = num
multiplies.append(num_string)
multiplies.append(num_string + "s")
# 2 and 3/4
for c in fraction_marker:
components = text.split(c)
if len(components) == 2:
# ensure first is not a fraction and second is a fraction
num1 = extractnumber_en(components[0])
num2 = extractnumber_en(components[1])
if num1 is not None and num2 is not None \
and num1 >= 1 and 0 < num2 < 1:
return num1 + num2
# 2 point 5
for c in decimal_marker:
components = text.split(c)
if len(components) == 2:
number = extractnumber_en(components[0])
decimal = extractnumber_en(components[1])
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))
aWords = text.split()
aWords = [word for word in aWords if word not in ["the", "a", "an"]]
val = False
prev_val = None
to_sum = []
for idx, word in enumerate(aWords):
if not word:
continue
prev_word = aWords[idx - 1] if idx > 0 else ""
next_word = aWords[idx + 1] if idx + 1 < len(aWords) else ""
# is this word already a number ?
if is_numeric(word):
# if word.isdigit(): # doesn't work with decimals
val = float(word)
# is this word the name of a number ?
if word in string_num_en:
val = string_num_en[word]
# is the prev word an ordinal number and current word is one?
# second one, third one
if ordinals and prev_word in string_num_ordinal_en and val is 1:
val = prev_val
# is the prev word a number and should we sum it?
# twenty two, fifty six
if prev_word in sums and word in string_num_en:
if val and val < 10:
val = prev_val + val
# is the prev word a number and should we multiply it?
# twenty hundred, six hundred
if word in multiplies:
if not prev_val:
prev_val = 1
val = prev_val * val
# is this a spoken fraction?
# half cup
if val is False:
val = isFractional_en(word, short_scale=short_scale)
# 2 fifths
if not ordinals:
next_value = isFractional_en(next_word, short_scale=short_scale)
if next_value:
if not val:
val = 1
val = val * next_value
# is this a negative number?
if val and prev_word and prev_word in negatives:
val = 0 - val
# let's make sure it isn't a fraction
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
else:
prev_val = val
# handle long numbers
# six hundred sixty six
# two million five hundred thousand
if word in multiplies and next_word not in multiplies:
to_sum.append(val)
val = 0
prev_val = 0
if val is not None:
for v in to_sum:
val = val + v
return val
def extractnumber_it(text, short_scale=False, 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
"""
string_num_ordinal_it = {}
# first, second...
if ordinals:
if short_scale:
for num in SHORT_ORDINAL_STRING_IT:
num_string = SHORT_ORDINAL_STRING_IT[num]
string_num_ordinal_it[num_string] = num
STRING_NUM_ITA[num_string] = num
else:
for num in LONG_ORDINAL_STRING_IT:
num_string = LONG_ORDINAL_STRING_IT[num]
string_num_ordinal_it[num_string] = num
STRING_NUM_ITA[num_string] = num
# negate next number (-2 = 0 - 2)
negatives = ['meno'] # 'negativo' non è usuale in italiano
# multiply the previous number (one hundred = 1 * 100)
multiplies = [
'decina', 'decine', 'dozzina', 'dozzine', 'centinaia', 'centinaio',
'migliaia', 'migliaio', 'mila'
]
# split sentence parse separately and sum ( 2 and a half = 2 + 0.5 )
fraction_marker = [' e ']
# decimal marker ( 1 point 5 = 1 + 0.5)
decimal_marker = [' punto ', ' virgola ']
if short_scale:
for num in SHORT_SCALE_IT:
num_string = SHORT_SCALE_IT[num]
STRING_NUM_ITA[num_string] = num
multiplies.append(num_string)
else:
for num in LONG_SCALE_IT:
num_string = LONG_SCALE_IT[num]
STRING_NUM_ITA[num_string] = num
multiplies.append(num_string)
# 2 e 3/4 ed altri casi
for separator in fraction_marker:
components = text.split(separator)
zeros = 0
if len(components) == 2:
# count zeros in fraction part
sub_components = components[1].split(' ')
for element in sub_components:
if element == 'zero' or element == '0':
zeros += 1
else:
break
# ensure first is not a fraction and second is a fraction
num1 = extractnumber_it(components[0])
num2 = extractnumber_it(components[1])
if num1 is not None and num2 is not None \
and num1 >= 1 and 0 < num2 < 1:
return num1 + num2
# sette e quaranta sette e zero zero due
elif num1 is not None and num2 is not None \
and num1 >= 1 and num2 > 1:
return num1 + num2 / pow(10, len(str(num2)) + zeros)
# 2 punto 5
for separator in decimal_marker:
zeros = 0
# count zeros in fraction part
components = text.split(separator)
if len(components) == 2:
sub_components = components[1].split(' ')
for element in sub_components:
if element == 'zero' or element == '0':
zeros += 1
else:
break
number = int(extractnumber_it(components[0]))
decimal = int(extractnumber_it(components[1]))
if number is not None and decimal is not None:
if '.' not in str(decimal):
return number + decimal / pow(10,
len(str(decimal)) + zeros)
all_words = text.split()
val = False
prev_val = None
to_sum = []
for idx, word in enumerate(all_words):
if not word:
continue
prev_word = all_words[idx - 1] if idx > 0 else ''
next_word = all_words[idx + 1] if idx + 1 < len(all_words) else ''
# is this word already a number ?
if is_numeric(word):
val = float(word)
# is this word the name of a number ?
if word in STRING_NUM_ITA:
val = STRING_NUM_ITA[word]
# tre quarti un quarto trenta secondi
if isFractional_it(word) and prev_val:
if word[:-1] == 'second' and not ordinals:
val = prev_val * 2
else:
val = prev_val
# is the prev word a number and should we multiply it?
# twenty hundred, six hundred
if word in multiplies:
if not prev_val:
prev_val = 1
val = prev_val * val
# is this a spoken fraction?
# mezza tazza
if val is False:
val = isFractional_it(word, short_scale=short_scale)
# 2 quinti
if not ordinals:
next_value = isFractional_it(next_word, short_scale=short_scale)
if next_value:
if not val:
val = 1
val = val * next_value
# is this a negative number?
if val and prev_word and prev_word in negatives:
val = 0 - val
if not val:
val = extractnumber_long_it(word)
# let's make sure it isn't a fraction
if not val:
# look for fractions like '2/3'
all_pieces = word.split('/')
if look_for_fractions(all_pieces):
val = float(all_pieces[0]) / float(all_pieces[1])
else:
prev_val = val
# handle long numbers
# six hundred sixty six
# two million five hundred thousand
if word in multiplies and next_word not in multiplies:
to_sum.append(val)
val = 0
prev_val = 0
elif extractnumber_long_it(word) > 100 and \
extractnumber_long_it(next_word) and \
next_word not in multiplies:
to_sum.append(val)
val = 0
prev_val = 0
if val is not None:
for addend in to_sum:
val = val + addend
return val
def extractnumber_sv(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
"""
aWords = text.split()
and_pass = False
valPreAnd = False
val = False
count = 0
while count < len(aWords):
word = aWords[count]
if is_numeric(word):
val = float(word)
elif word == "första":
val = 1
elif word == "andra":
val = 2
elif word == "tredje":
val = 3
elif word == "fjärde":
val = 4
elif word == "femte":
val = 5
elif word == "sjätte":
val = 6
elif is_fractional_sv(word):
val = is_fractional_sv(word)
else:
if word == "en":
val = 1
if word == "ett":
val = 1
elif word == "två":
val = 2
elif word == "tre":
val = 3
elif word == "fyra":
val = 4
elif word == "fem":
val = 5
elif word == "sex":
val = 6
elif word == "sju":
val = 7
elif word == "åtta":
val = 8
elif word == "nio":
val = 9
elif word == "tio":
val = 10
if val:
if count < (len(aWords) - 1):
wordNext = aWords[count + 1]
else:
wordNext = ""
valNext = is_fractional_sv(wordNext)
if valNext:
val = val * valNext
aWords[count + 1] = ""
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
elif and_pass:
# added to value, quit here
val = valPreAnd
break
else:
count += 1
continue
aWords[count] = ""
if and_pass:
aWords[count - 1] = '' # remove "och"
val += valPreAnd
elif count + 1 < len(aWords) and aWords[count + 1] == 'och':
and_pass = True
valPreAnd = val
val = False
count += 2
continue
elif count + 2 < len(aWords) and aWords[count + 2] == 'och':
and_pass = True
valPreAnd = val
val = False
count += 3
continue
break
if not val:
return False
return val
def extractnumber_it(text, short_scale=False, 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
"""
string_num_ordinal_it = {}
# first, second...
if ordinals:
if short_scale:
for num in SHORT_ORDINAL_STRING_IT:
num_string = SHORT_ORDINAL_STRING_IT[num]
string_num_ordinal_it[num_string] = num
STRING_NUM_ITA[num_string] = num
else:
for num in LONG_ORDINAL_STRING_IT:
num_string = LONG_ORDINAL_STRING_IT[num]
string_num_ordinal_it[num_string] = num
STRING_NUM_ITA[num_string] = num
# negate next number (-2 = 0 - 2)
negatives = ['meno'] # 'negativo' non è usuale in italiano
# multiply the previous number (one hundred = 1 * 100)
multiplies = ['decina', 'decine', 'dozzina', 'dozzine',
'centinaia', 'centinaio', 'migliaia', 'migliaio', 'mila']
# split sentence parse separately and sum ( 2 and a half = 2 + 0.5 )
fraction_marker = [' e ']
# decimal marker ( 1 point 5 = 1 + 0.5)
decimal_marker = [' punto ', ' virgola ']
if short_scale:
for num in SHORT_SCALE_IT:
num_string = SHORT_SCALE_IT[num]
STRING_NUM_ITA[num_string] = num
multiplies.append(num_string)
else:
for num in LONG_SCALE_IT:
num_string = LONG_SCALE_IT[num]
STRING_NUM_ITA[num_string] = num
multiplies.append(num_string)
# 2 e 3/4 ed altri casi
for separator in fraction_marker:
components = text.split(separator)
zeros = 0
if len(components) == 2:
# count zeros in fraction part
sub_components = components[1].split(' ')
for element in sub_components:
if element == 'zero' or element == '0':
zeros += 1
else:
break
# ensure first is not a fraction and second is a fraction
num1 = extractnumber_it(components[0])
num2 = extractnumber_it(components[1])
if num1 is not None and num2 is not None \
and num1 >= 1 and 0 < num2 < 1:
return num1 + num2
# sette e quaranta sette e zero zero due
elif num1 is not None and num2 is not None \
and num1 >= 1 and num2 > 1:
return num1 + num2 / pow(10, len(str(num2)) + zeros)
# 2 punto 5
for separator in decimal_marker:
zeros = 0
# count zeros in fraction part
components = text.split(separator)
if len(components) == 2:
sub_components = components[1].split(' ')
for element in sub_components:
if element == 'zero' or element == '0':
zeros += 1
else:
break
number = int(extractnumber_it(components[0]))
decimal = int(extractnumber_it(components[1]))
if number is not None and decimal is not None:
if '.' not in str(decimal):
return number + decimal / pow(10,
len(str(decimal)) + zeros)
all_words = text.split()
val = False
prev_val = None
to_sum = []
for idx, word in enumerate(all_words):
if not word:
continue
prev_word = all_words[idx - 1] if idx > 0 else ''
next_word = all_words[idx + 1] if idx + 1 < len(all_words) else ''
# is this word already a number ?
if is_numeric(word):
val = float(word)
# is this word the name of a number ?
if word in STRING_NUM_ITA:
val = STRING_NUM_ITA[word]
# tre quarti un quarto trenta secondi
if isFractional_it(word) and prev_val:
if word[:-1] == 'second' and not ordinals:
val = prev_val * 2
else:
val = prev_val
# is the prev word a number and should we multiply it?
# twenty hundred, six hundred
if word in multiplies:
if not prev_val:
prev_val = 1
val = prev_val * val
# is this a spoken fraction?
# mezza tazza
if val is False:
val = isFractional_it(word, short_scale=short_scale)
# 2 quinti
if not ordinals:
next_value = isFractional_it(next_word, short_scale=short_scale)
if next_value:
if not val:
val = 1
val = val * next_value
# is this a negative number?
if val and prev_word and prev_word in negatives:
val = 0 - val
if not val:
val = extractnumber_long_it(word)
# let's make sure it isn't a fraction
if not val:
# look for fractions like '2/3'
all_pieces = word.split('/')
if look_for_fractions(all_pieces):
val = float(all_pieces[0]) / float(all_pieces[1])
else:
prev_val = val
# handle long numbers
# six hundred sixty six
# two million five hundred thousand
if word in multiplies and next_word not in multiplies:
to_sum.append(val)
val = 0
prev_val = 0
elif extractnumber_long_it(word) > 100 and \
extractnumber_long_it(next_word) and \
next_word not in multiplies:
to_sum.append(val)
val = 0
prev_val = 0
if val is not None:
for addend in to_sum:
val = val + addend
return val
def extractnumber_en(text):
"""
This function prepares the given text for parsing by making
numbers consistent, getting rid of contractions, etc.
Args:
text (str): the string to normalize
Returns:
(int) or (float): The value of extracted number
"""
aWords = text.split()
aWords = [word for word in aWords if word not in ["the", "a", "an"]]
and_pass = False
valPreAnd = False
val = False
count = 0
while count < len(aWords):
word = aWords[count]
if is_numeric(word):
# if word.isdigit(): # doesn't work with decimals
val = float(word)
elif word == "first":
val = 1
elif word == "second":
val = 2
elif isFractional_en(word):
val = isFractional_en(word)
else:
if word == "one":
val = 1
elif word == "two":
val = 2
elif word == "three":
val = 3
elif word == "four":
val = 4
elif word == "five":
val = 5
elif word == "six":
val = 6
elif word == "seven":
val = 7
elif word == "eight":
val = 8
elif word == "nine":
val = 9
elif word == "ten":
val = 10
if val:
if count < (len(aWords) - 1):
wordNext = aWords[count + 1]
else:
wordNext = ""
valNext = isFractional_en(wordNext)
if valNext:
val = val * valNext
aWords[count + 1] = ""
# if val == False:
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
# if (len(aPieces) == 2 and is_numeric(aPieces[0])
# and is_numeric(aPieces[1])):
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
elif and_pass:
# added to value, quit here
val = valPreAnd
break
else:
count += 1
continue
aWords[count] = ""
if and_pass:
aWords[count - 1] = '' # remove "and"
val += valPreAnd
elif count + 1 < len(aWords) and aWords[count + 1] == 'and':
and_pass = True
valPreAnd = val
val = False
count += 2
continue
elif count + 2 < len(aWords) and aWords[count + 2] == 'and':
and_pass = True
valPreAnd = val
val = False
count += 3
continue
break
# if val == False:
if not val:
return False
# Return the string with the number related words removed
# (now empty strings, so strlen == 0)
aWords = [word for word in aWords if len(word) > 0]
text = ' '.join(aWords)
return val
def _extract_whole_number_with_text_en(tokens, short_scale, ordinals):
"""
Handle numbers not handled by the decimal or fraction functions. This is
generally whole numbers. Note that phrases such as "one half" will be
handled by this function, while "one and a half" are handled by the
fraction function.
Args:
tokens [_Token]:
short_scale boolean:
ordinals boolean:
Returns:
int or float, [_Tokens]
The value parsed, and tokens that it corresponds to.
"""
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):
current_val = None
if next_val:
next_val = None
continue
word = token.word
if word in _ARTICLES or word in _NEGATIVES:
number_words.append(token)
continue
prev_word = tokens[idx - 1].word if idx > 0 else ""
next_word = tokens[idx + 1].word if idx + 1 < len(tokens) else ""
if word not in string_num_scale and \
word not in _STRING_NUM_EN and \
word not in _SUMS and \
word not in multiplies and \
not (ordinals and word in string_num_ordinal) and \
not is_numeric(word) and \
not isFractional_en(word, short_scale=short_scale) and \
not look_for_fractions(word.split('/')):
words_only = [token.word for token in number_words]
if number_words and not all(
[w in _ARTICLES | _NEGATIVES for w in words_only]):
break
else:
number_words = []
continue
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 prev_word not in _NEGATIVES \
and prev_word not in _ARTICLES:
number_words = [token]
elif prev_word in _SUMS and word in _SUMS:
number_words = [token]
else:
number_words.append(token)
# is this word already a number ?
if is_numeric(word):
if word.isdigit(): # doesn't work with decimals
val = int(word)
else:
val = float(word)
current_val = val
# is this word the name of a number ?
if word in _STRING_NUM_EN:
val = _STRING_NUM_EN.get(word)
current_val = val
elif word in string_num_scale:
val = string_num_scale.get(word)
current_val = val
elif ordinals and word in string_num_ordinal:
val = string_num_ordinal[word]
current_val = val
# is the prev word an ordinal number and current word is one?
# second one, third one
if ordinals and prev_word in string_num_ordinal and val is 1:
val = prev_val
# is the prev word a number and should we sum it?
# twenty two, fifty six
if prev_word in _SUMS and val and val < 10:
val = prev_val + val
# is the prev word a number and should we multiply it?
# twenty hundred, six hundred
if word in multiplies:
if not prev_val:
prev_val = 1
val = prev_val * val
# is this a spoken fraction?
# half cup
if val is False:
val = isFractional_en(word, short_scale=short_scale)
current_val = val
# 2 fifths
if not ordinals:
next_val = isFractional_en(next_word, short_scale=short_scale)
if next_val:
if not val:
val = 1
val = val * next_val
number_words.append(tokens[idx + 1])
# is this a negative number?
if val and prev_word and prev_word in _NEGATIVES:
val = 0 - val
# let's make sure it isn't a fraction
if not val:
# look for fractions like "2/3"
aPieces = word.split('/')
if look_for_fractions(aPieces):
val = float(aPieces[0]) / float(aPieces[1])
current_val = val
else:
if prev_word in _SUMS and word not in _SUMS and current_val >= 10:
# Backtrack - we've got numbers we can't sum.
number_words.pop()
val = prev_val
break
prev_val = val
# handle long numbers
# six hundred sixty six
# two million five hundred thousand
if word in multiplies and next_word not in multiplies:
to_sum.append(val)
val = 0
prev_val = 0
if val is not None and to_sum:
val += sum(to_sum)
return val, number_words
