2

This is a repeat of a previous question from several years ago that did not receive an answer.

Oracle has a string matching package, UTL_MATCH, which contains the subprogram JARO_WINKLER_SIMILARITY.

Jaro Winkler Similarity is a variant of the Jaro Distance, but includes a prefix scale factor to account for common characters at the start of two strings being matched. I cannot find any documentation for the default prefix scale in the UTL_MATCH package, or how to set this manually. Does anyone know of any documentation on this?

Improve this question
1

You are right that the Oracle documentation does not mention a prefix scale factor. This leads me to conclude that what Oracle is offering is a simpler version of the formula where there is no option to choose a prefix length. There is no other Oracle documentation on this that I know of and it looks unchanged from 10g when it was first implemented.

Even if there was a hidden implementation it would not be advisable to use it in production code as there is no guarantee of support or that it will remain available.

From the Wikipedia reference you gave it looks like you could create your own function which implements the Jaro Winkler similarity formula exactly.

Improve this answer
2
  • Thanks. I am not sure what you mean by 'not implemented' - the package is certainly functioning and providing different results to my own implementation of the Jaro Distance metric (also checked manually), which leads me to believe there must be a prefix scale involved. I suppose it may be possible to reverse engineer the prefix scale though, I will look into that too. – Nick Jun 5 '19 at 16:47
  • 1
    Update: through trial-and-error I managed to determine that the default prefix scale factor in the JARO_WINKLER_SIMILARITY function equals 0.1. This is consistent with the standard value used by Winkler, which is encouraging. Unfortunately, it seems that @kevinsky is correct in there not being a way to choose a prefix scale manually. – Nick Jun 5 '19 at 17:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.