Written by Imran
on September 17, 2021

Fuzzy word replace from string in Python

2 min read

Have you ever faced problems like the following?

You are given a string, for example "Apple MacBook Pro appl". Now you have to remove the word "Apple" from the string. You can easily use python’s default replace() function to do that. It’ll only match and replace the words that exactly matches to Apple. But it is required that you also remove the words that might be misspelled, like appl. How can we do that?

Fuzzy string matching techniques come into play here.

alt text

We can solve this problem in 2 ways. One is through using Python’s default difflib library, and the other one is through using rapidfuzz, a fuzzy string matching library.

Below I’ve given codes for both of the approaches.

Using difflib:

import difflib

def fuzzy_replace(word, string):
    '''
    Returns the cleaned string.

            Parameters:
                    word (str): word to be replaced
                    string (str): given string from where we should remove the word

            Returns:
                    string (str): the cleaned string.
    '''
    tokens = string.split()
    matches = difflib.get_close_matches(word=word, possibilities=tokens, n=len(tokens), cutoff=0.6)
    return ' '.join(filter(lambda x: x not in matches, tokens))

Using rapidfuzz:

To use rapidfuzz, we need to install it using the following command

pip install rapidfuzz

Now, let’s implement the function using rapidfuzz.

from rapidfuzz import process, fuzz

def fuzzy_replace(word, string):
    '''
    Returns the cleaned product name as string.

            Parameters:
                    word (str): brand name as string
                    string (str): product name as string

            Returns:
                    product name (str): the cleaned product name as string.
    '''
    tokens = string.split()
    matches = process.extract(query=word, choices=tokens, scorer=fuzz.WRatio, score_cutoff=60.0, limit=len(tokens))
    idx_matches = [m[2] for m in matches]
    return ' '.join(filter(lambda token_idx: tokens.index(token_idx) not in idx_matches, tokens))

To learn more about rapidfuzz’s functions please visit here

Now, if we use the function we can see the output as follows:

word = "Apple"
string = "Apple MacBook Pro appl"

print(fuzzy_replace(word, string))

Output: MacBook Pro

Miscellaneous

← → Top