[
https://issues.apache.org/jira/browse/TEXT-191?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17317852#comment-17317852
]
Bradley Rumball commented on TEXT-191:
--------------------------------------
Hi [~kinow]
No worries at all, I came here due to using this function on a project and
finding that the distance was driving one of the fitness functions I was
working on in the wrong direction; I think it's safe to say it's a bug that
needed squashing!
Thanks for updating the PR to using 1-similarity for the distance; this is far
superior; I wanted to do this but wasn't sure if it was previously implemented
like this to reduce stack depth; in short, I'm very much in favour of what
you've done.
In the future, it would be great to see the package of "similarity" classes
provided as final utility classes; it's unnatural to have to create instances
for them.
I'd happily work on this and provide a PR for in a future version; it would be
great to hear everyone's thoughts!
Aside from this, I'm happy for the merge to go ahead!
Best wishes,
Bradley
> JaroWinklerDistance returns the same values as JaroWinklerSimilarity
> --------------------------------------------------------------------
>
> Key: TEXT-191
> URL: https://issues.apache.org/jira/browse/TEXT-191
> Project: Commons Text
> Issue Type: Bug
> Affects Versions: 1.9
> Environment: Windows 10
> AdoptOpenJDK 11
> Reporter: Alexander Foerster
> Assignee: Bruno P. Kinoshita
> Priority: Major
> Labels: JaroWinkler, easyfix, newbie
> Fix For: 1.10
>
>
>
> The following two code snippets give the same output:
> {code:java}
> // checking similarity
> JaroWinklerSimilarity jws = new JaroWinklerSimilarity();
> System.out.println("SIMILARITY: " + jws.apply("foo", "foo") + " " +
> jws.apply("foo", "bar"));
> // SIMILARITY: 1.0 0.0{code}
> {code:java}
> // checking distance
> JaroWinklerDistance jwd = new JaroWinklerDistance();
> System.out.println("DISTANCE: " + jwd.apply("foo", "foo") + " " +
> jwd.apply("foo", "bar"));
> // DISTANCE: 1.0 0.0
> {code}
> I would expect the distance to be 1 - similarity, but as you can see distance
> and similarity return the same values.
>
> An easy fix would be to use the commented lines in JaroWinklerDistance.java:
> {code:java}
> // JaroWinklerSimilarity similarity = new JaroWinklerSimilarity();
> // return 1 - similarity.apply(left, right);
> {code}
> and maybe make _similarity_ be a final property/variable.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
