feat/number_spans

feat/normalize_decimals

port lingua_nostra/pull/20 - support decimal markers

rebase of MycroftAI#69

Co-authored-by: jarbasal <[email protected]>

Author: JarbasAl

lingua_franca/lang/common_data_en.py

@@ -247,7 +247,7 @@
 
 
 # negate next number (-2 = 0 - 2)
-_NEGATIVES_EN = {"negative", "minus"}
+_NEGATIVES_EN = {"negative", "minus", "-"}
 
 # sum the next number (twenty two = 20 + 2)
 _SUMS_EN = {'twenty', '20', 'thirty', '30', 'forty', '40', 'fifty', '50',

lingua_franca/lang/parse_common.py

@@ -192,6 +192,18 @@ def normalize(self, utterance="", remove_articles=None):
         return utterance
 
 
+def normalize_decimals(text, decimal):
+    """
+        Replace 'decimal' with decimal periods so Python can floatify them
+    """
+    regex = r"\b\d+" + decimal + r"{1}\d+\b"
+    sanitize_decimals = re.compile(regex)
+    for _, match in enumerate(re.finditer(sanitize_decimals, text)):
+        text = text.replace(match.group(
+            0), match.group(0).replace(decimal, '.'))
+    return text
+
+
 def match_yes_or_no(text, lang):
     resource_file = resolve_resource_file(f"text/{lang}/yesno.json")
     if not resource_file:

lingua_franca/lang/parse_cs.py

@@ -23,7 +23,7 @@
     _LONG_ORDINAL_CS, _LONG_SCALE_CS, _SHORT_SCALE_CS, _SHORT_ORDINAL_CS, \
     _FRACTION_STRING_CS, _MONTHS_CONVERSION, _MONTHS_CZECH, _TIME_UNITS_CONVERSION, \
     _ORDINAL_BASE_CS  # _ARTICLES_CS
-
+from lingua_franca.lang.parse_common import normalize_decimals
 import re
 import json
 from lingua_franca import resolve_resource_file
@@ -579,7 +579,7 @@ def _initialize_number_data(short_scale):
     return multiplies, string_num_ordinal_cs, string_num_scale_cs
 
 
-def extract_number_cs(text, short_scale=True, ordinals=False):
+def extract_number_cs(text, short_scale=True, ordinals=False, decimal='.'):
     """
     This function extracts a number from a text string,
     handles pronunciations in long scale and short scale
@@ -590,11 +590,17 @@ def extract_number_cs(text, short_scale=True, ordinals=False):
         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
+        decimal (str): character to use as decimal point. defaults to '.'
     Returns:
         (int) or (float) or False: The extracted number or False if no number
                                    was found
+    Note:
+        will always extract numbers formatted with a decimal dot/full stop,
+        such as '3.5', even if 'decimal' is specified.
 
     """
+    if decimal != '.':
+        text = normalize_decimals(text, decimal)
     return _extract_number_with_text_cs(tokenize(text.lower()),
                                         short_scale, ordinals).value
 
@@ -1560,20 +1566,25 @@ def isFractional_cs(input_str, short_scale=True):
     return False
 
 
-def extract_numbers_cs(text, short_scale=True, ordinals=False):
+def extract_numbers_cs(text, short_scale=True, ordinals=False, decimal='.'):
     """
         Takes in a string and extracts a list of numbers.
 
     Args:
-        text (str): the string to extract a number from
-        short_scale (bool): Use "short scale" or "long scale" for large
-            numbers -- over a million.  The default is short scale, which
-            is now common in most English speaking countries.
-            See https://en.wikipedia.org/wiki/Names_of_large_numbers
-        ordinals (bool): consider ordinal numbers, e.g. third=3 instead of 1/3
+        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
+        decimal (str): character to use as decimal point. defaults to '.'
     Returns:
-        list: list of extracted numbers as floats
+        (int) or (float) or False: The extracted number or False if no number
+                                   was found
+    Note:
+        will always extract numbers formatted with a decimal dot/full stop,
+        such as '3.5', even if 'decimal' is specified.
+
     """
+    if decimal != '.':
+        text = normalize_decimals(text, decimal)
     results = _extract_numbers_with_text_cs(tokenize(text),
                                             short_scale, ordinals)
     return [float(result.value) for result in results]

lingua_franca/lang/parse_da.py

@@ -20,22 +20,31 @@
 from lingua_franca.lang.common_data_da import _DA_NUMBERS
 from lingua_franca.lang.format_da import pronounce_number_da
 from lingua_franca.time import now_local
+from lingua_franca.lang.parse_common import normalize_decimals
 
 
-def extract_number_da(text, short_scale=True, ordinals=False):
+def extract_number_da(text, short_scale=True, ordinals=False, decimal='.'):
     """
-    This function prepares the given text for parsing by making
-    numbers consistent, getting rid of contractions, etc.
+    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
+        decimal (str): character to use as decimal point. defaults to '.'
     Returns:
-        (int) or (float): The value of extracted number
-
-
-    undefined articles cannot be suppressed in German:
-    'ein Pferd' means 'one horse' and 'a horse'
+        (int) or (float) or False: The extracted number or False if no number
+                                   was found
+    Note:
+        will always extract numbers formatted with a decimal dot/full stop,
+        such as '3.5', even if 'decimal' is specified.
 
     """
+    if decimal != '.':
+        text = normalize_decimals(text, decimal)
     # TODO: short_scale and ordinals don't do anything here.
     # The parameters are present in the function signature for API compatibility
     # reasons.
@@ -869,20 +878,25 @@ def normalize_da(text, remove_articles=True):
     return normalized[1:]  # strip the initial space
 
 
-def extract_numbers_da(text, short_scale=True, ordinals=False):
+def extract_numbers_da(text, short_scale=True, ordinals=False, decimal='.'):
     """
         Takes in a string and extracts a list of numbers.
 
-    Args:
-        text (str): the string to extract a number from
-        short_scale (bool): Use "short scale" or "long scale" for large
-            numbers -- over a million.  The default is short scale, which
-            is now common in most English speaking countries.
-            See https://en.wikipedia.org/wiki/Names_of_large_numbers
-        ordinals (bool): consider ordinal numbers, e.g. third=3 instead of 1/3
+     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
+        decimal (str): character to use as decimal point. defaults to '.'
     Returns:
-        list: list of extracted numbers as floats
+        (int) or (float) or False: The extracted number or False if no number
+                                   was found
+    Note:
+        will always extract numbers formatted with a decimal dot/full stop,
+        such as '3.5', even if 'decimal' is specified.
+
     """
+    if decimal != '.':
+        text = normalize_decimals(text, decimal)
     return extract_numbers_generic(text, pronounce_number_da, extract_number_da,
                                    short_scale=short_scale, ordinals=ordinals)
 

lingua_franca/lang/parse_de.py

@@ -21,6 +21,7 @@
 from lingua_franca.lang.common_data_de import _DE_NUMBERS
 from lingua_franca.lang.format_de import pronounce_number_de
 from lingua_franca.time import now_local
+from lingua_franca.lang.parse_common import normalize_decimals
 
 
 de_numbers = {
@@ -143,20 +144,28 @@ def repl(match):
     return (duration, text)
 
 
-def extract_number_de(text, short_scale=True, ordinals=False):
+def extract_number_de(text, short_scale=True, ordinals=False, decimal='.'):
     """
-    This function prepares the given text for parsing by making
-    numbers consistent, getting rid of contractions, etc.
+    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
+        decimal (str): character to use as decimal point. defaults to '.'
     Returns:
-        (int) or (float): The value of extracted number
-
-
-    undefined articles cannot be suppressed in German:
-    'ein Pferd' means 'one horse' and 'a horse'
+        (int) or (float) or False: The extracted number or False if no number
+                                   was found
+    Note:
+        will always extract numbers formatted with a decimal dot/full stop,
+        such as '3.5', even if 'decimal' is specified.
 
     """
+    if decimal != '.':
+        text = normalize_decimals(text, decimal)
     # TODO: short_scale and ordinals don't do anything here.
     # The parameters are present in the function signature for API compatibility
     # reasons.
@@ -1003,20 +1012,28 @@ def normalize_de(text, remove_articles=True):
     return normalized[1:]  # strip the initial space
 
 
-def extract_numbers_de(text, short_scale=True, ordinals=False):
-    """
-        Takes in a string and extracts a list of numbers.
-
-    Args:
-        text (str): the string to extract a number from
-        short_scale (bool): Use "short scale" or "long scale" for large
-            numbers -- over a million.  The default is short scale, which
-            is now common in most English speaking countries.
-            See https://en.wikipedia.org/wiki/Names_of_large_numbers
-        ordinals (bool): consider ordinal numbers, e.g. third=3 instead of 1/3
-    Returns:
-        list: list of extracted numbers as floats
+def extract_numbers_de(text, short_scale=True, ordinals=False, decimal='.'):
     """
+       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
+           decimal (str): character to use as decimal point. defaults to '.'
+       Returns:
+           (int) or (float) or False: The extracted number or False if no number
+                                      was found
+       Note:
+           will always extract numbers formatted with a decimal dot/full stop,
+           such as '3.5', even if 'decimal' is specified.
+
+       """
+    if decimal != '.':
+        text = normalize_decimals(text, decimal)
     return extract_numbers_generic(text, pronounce_number_de, extract_number_de,
                                    short_scale=short_scale, ordinals=ordinals)