Skip to content

DoubleBite/Numbers-from-String

Repository files navigation

Numbers from String

This Python module provides functions that get the numbers or the numeric string tokens in an input string.

To capture the numerals in a piece of text is a common preprocess for retrieving numerical information from documents. However, due to the various representations of these numerals, it's somewhat tricky to capture them all using simple rules. We packed several regex rules with comprehensive coverage in this library and hope it can be a useful tool for NLP researchers.

Installation

pip install nums_from_string

Usage

  1. Extract numbers from a string
>>> string1 = "U.S. goods and services trade with China totaled an estimated $710.4 billion in 2017. "
>>> nums_from_string.get_nums(string1)
[710.4, 2017]

>>> string2 = "David spent .25 billion dollars buying a building and 600,000.5 dollars getting himself a car."
>>> nums_from_string.get_nums(string2)
[0.25, 600000.5]
  1. Extract numeric strings from a string
>>> string1 = "U.S. goods and services trade with China totaled an estimated $710.4 billion in 2017. "
>>> nums_from_string.get_numeric_string_tokens(string1)
['710.4', '2017']

>>> string2 = "David spent .25 billion dollars buying a building and 600,000.5 dollars getting himself a car."
>>> nums_from_string.get_numeric_string_tokens(string2)
['.25', '600,000.5']

>>> string3 = "Find the product of 4 and -5."
>>> nums_from_string.get_numeric_string_tokens(string3)
['4', '-5']

>>> string4 = "The flight number is Airbus A330-300"
>>> nums_from_string.get_numeric_string_tokens(string4, no_minus=True)
['330', '300']
  1. Convert strings to numbers
>>> s0 = "255"
>>> nums_from_string.to_num(s0)
255

>>> s1 = "-255,000.0"
>>> nums_from_string.to_num(s1)
-255000.0

>>> s2 = "87/25"
>>> nums_from_string.to_num(s2)
Fraction(87, 25)

>>> s3 = "a1b2"
>>> nums_from_string.to_num(s3)
Traceback (most recent call last):
    ...
ValueError: Invalid numerical string!

Todo

  • Capture the pattern of fractions in a string
  • Capture the patterns like this "-3.5/11"

Reference

License

This project is licensed under the terms of the MIT license.

Releases

No releases published

Packages

No packages published