how fuzzywuzzy python works - when there is no matching words between sentences?

Question

I am using fuzzywuzzy match to find similarity between sentences.

when I compare these two sentences- 'user attempts login' and 'acceptance criteria'

fuzz.token_set_ratio('user attempts login', 'acceptance criteria')

it gives me a score of 42 .

could someone please help me understand how we get score of 42 when there are no matching words ??

Answer 1

Steps of the Algorithm

Token_set_ratio performs the following steps:

1. split sentence and remove duplicates
2. create three lists of
   • remainder1 = words that are only in the first sentence
   • remainder2 = words that are only in the second sentence
   • intersection = words that are in both sentences
3. sort the words in the three lists and join the elements to a combined string
   • sorted_remainder1
   • sorted_remainder2
   • sorted_intersection
4. join the strings in the following way:
   • combined1 = <sorted_intersection><sorted_remainder1>
   • combined2 = <sorted_intersection><sorted_remainder2>
5. calculate the following similarities:
   • fuzz.ratio(sorted_intersection, combined1)
   • fuzz.ratio(sorted_intersection, combined2)
   • fuzz.ratio(combined1, combined2)
6. return the maximum of those similarities

Example

For the strings user attempts login and acceptance criteria this leads to the following result:

remainder1 = ['user', 'attempts', 'login']
remainder2 = ['acceptance', 'criteria']
intersection = []
sorted_remainder1 = 'attempts login user'
sorted_remainder2 = 'acceptance criteria'
combined1 = 'attempts login user'
combined2 = 'acceptance criteria'

fuzz.ratio(sorted_intersection, combined1) = 0
fuzz.ratio(sorted_intersection, combined2) = 0
fuzz.ratio(combined1, combined2) = 42

In your specific case this is a similar result to fuzz.token_sort_ratio, which only sorts the words in both sentences and compares them using fuzz.ratio.