🧞 ✍🏽 🧝🏻 Using the FuzzyWuzzy library for fuzzy comparisons in Python. Levenshtein Distance (editorial distance) 🤶🏼 🍧 🏔️

He was not sure that he heard correctly. So much depended on it! But do not ask again? (c) Boris Akunin. The whole world is a theater.

While working on the voice assistant that was mentioned in the previous article , I realized that I just can’t help but share the beautiful FuzzyWuzzy library with you .

In short, thanks to her, it is possible to make a fuzzy string comparison without any suffering.

First steps

To get started, you need to do two steps:

/ IMPORTANT! Python version 2.7 and higher /

Step 1. Installation.

Open the command line and enter:

pip install fuzzywuzzy

Press Enter.
Next, install python-Levenshtein in the same way to speed up string matching by 3-10 times.

pip install python-Levenshtein

After installation is complete, the library is ready to import.

Step 2. Importing into the project.

from fuzzywuzzy import fuzz
from fuzzywuzzy import process

Functional

1. The most common comparison:

a = fuzz.ratio(' ', ' ')
print(a)
#  : 100

If we change a couple of characters, then the output will get a different number.

a = fuzz.ratio(' ', ' ')
print(a)
#  : 84

2. Partial comparison.

This type of comparison in the entire second line looks for a match with the initial one, for example:

a = fuzz.partial_ratio(' ', ' !')
print(a)
#  : 100

a = fuzz.partial_ratio(' ', ' ,  ')
print(a)
#  : 100

But you should remember about the register, since

a = fuzz.partial_ratio(' ', ' ,  ')
print(a) 

#  : 90

3.

Token comparison 1) Token Sort Ratio

Words are compared with each other, regardless of case or order

a = fuzz.token_sort_ratio('  ', '  ')
print(a)
#  : 100

a = fuzz.token_sort_ratio('  ', '   ')
print(a)
#  : 78

a = fuzz.token_sort_ratio('1 2   ', '1   2 ')
print(a)
#  : 100

2) Token Set Ratio
This comparison, unlike the past, equates strings, if their difference is the repetition of words.

a = fuzz.token_set_ratio('  ', '     ')
print(a)
#  : 100

4. Advanced regular comparison

In many cases, it is more advisable to use exactly WRatio , since it is case-sensitive and punctuation (not dividing the string)

a = fuzz.WRatio('  ', '!  !')
print(a)
#  : 100

a = fuzz.WRatio('  ', '!,  !')
print(a)
#  : 97

5. Working with the list

To compare the lines with the lines from the list, the process module is used

city = ["", "-", "", "", "", "", "", "", "", "", ""]
a = process.extract("", city, limit=2)
#  limit     5
print(a)
#  : [('', 100), ('', 62)]

If only the first one in the list is needed, then it is better to use extractOne

city = ["", "-", "", "", "", "", "", "", "", "", ""]
a = process.extractOne("", city)
print(a)
#  : ('', 90)

Application

How and where to apply all of the above is up to you, but here is an example from my term paper :

#    
try:
    files = os.listdir('C:\\Users\\hartp\\Desktop\\')
    filestart = process.extractOne(namerec, files)
    if filestart[1] >= 80:
        os.startfile('C:\\Users\\hartp\\Desktop\\' + filestart[0])
    else:
        speak('  ')
except FileNotFoundError:
    speak('  ')

Let's go over the code and understand what's what. With the os.listdir

command , we get a list of all the files that are present at the end of the specified path (in our case, to the desktop).

files = os.listdir('C:\\Users\\hartp\\Desktop\\')
print(files)
#  : 'Visual Studio 2019.lnk', 'Visual Studio Code.lnk', 'WarThunder.lnk', 'WpfApp14', 'Yandex.lnk', '.docx', '.txt', ' .txt'  ..

Next is a comparison of the lines of the file list with the name of the file that the user named (variable namerec ). I hope you noticed that the result of the extractOne function is a tuple of string and number (similarity index)

Example from last chapter

city = ["", "-", "", "", "", "", "", "", "", "", ""]
a = process.extractOne("", city)
print(a)
#  : ('', 90)

Based on this, we check the similarity index filestart [1]> = 80 ([1], since the tuple is numbered from 0, as in an array) and, if the condition is true, then run the os.startfile function with a file called filestart [0 ]. Otherwise, if the similarity index is less than 80 or an error occurs that the file was not found, we inform the user through the speak function .