Discrepancies in Scores when compared to basic implementations.

[kiran-vj]

I was comparing scores for this package against a basic Java implementation which can be found here https://www.geeksforgeeks.org/jaro-and-jaro-winkler-similarity/

While for majority of the cases I observed identical scores- but I also see cases where there are large differences.

For example:

s1 = "aidahaddad", s2 = "thadwindle";

The basic java version returns a score= 0.5666666666666667

While the python package returns a score= 0.6

All the parameters such as prefix_weight/max_prefix_len etc. seem to be in line with the python package.

What could be the root cause of such discrepancies? Are there any corresponding Java packages where we could observe scores identical to the python implementation here.