Jaro-Winkler returning the wrong value?

[wrh...@gmail.com]

I was doing some testing of the implementation of the Jaro-Winkler comparator in Duke and I think there may be an issue but I'm not 100% sure.

I first computed the JW distance using the apache commons library.  Then I did it with Duke.  There are several cases where they don't match.

I found an online app (link below) that matches 1:1 with Apache commons, so that would lead me to think Duke is wrong - but who knows (maybe they are using the same library)?

https://asecuritysite.com/forensics/simstring

Here are some examples where they differ:

GAGEL, MONTANEZ (AC=0.0, Duke=0.549)

KATHLEEN, LUCIA (AC=0.0, Duke=0.441)

HOOD, MONTANEZ (AC=0.458, Duke=0.583)

SILLE, ZENATNOM (AC=0.0, Duke=0.441)

HTINEK, ZENATNOM (AC=0.513, Duke=0.402)