Jaro-Winkler Distance Algorithm Match Value Question about natural HOT 13 CLOSED

Hey Damon!

You're right, it should return 1 when they are exact string matches. I think I may have implemented the "winkler" part of the algorithm incorrectly. Hopefully I can take a look at the issue this week (but I've switched priorities since I had a daughter a couple of months ago :-))

Adam

from natural.

Hey Mate - congrats on the birth of your daughter. I know what you mean - everything changes when you've got kids :)

I've got a better understanding of the algorithm now after working through it a little and hacking around with an implementation to try and help get the gist of it (https://github.com/DamonOehlman/strcmp.js). No idea if I've got it right though - it's really tough to find a reference implementation to compare against for testing purposes, and my implementation is a little different to most others I have seen...

If I get a little less snowed under in the next few days, I'll see if I can work out where the issue is in Natural's implementation...

from natural.

@adamb0mb congrats, man!

in the next couple of days i'll have a look as well. it may have even been me that broke it when trying to tie some things together.

from natural.

@chrisumbel I'm trying to get jasmine-node running for Natural on my machine, and I'm having issues with directories.

C:\git\natural>jasmine-node spec

module.js:340
throw err;
^
Error: Cannot find module 'lib/natural/tokenizers/aggressive_tokenizer'
at Function.Module._resolveFilename (module.js:338:15)
....
at Object. (C:\git\natural\spec\aggressive_tokenizer_spec.js:23:17)
...
at Module.require (module.js:362:17)

This is the current code:

var Tokenizer = require('lib/natural/tokenizers/aggressive_tokenizer')
  tokenizer = new Tokenizer();

I can get past this error message if I change the aggressive_tokenizer_spec.js to look for the library starting from the parent directory:

var Tokenizer = require('../lib/natural/tokenizers/aggressive_tokenizer')
  tokenizer = new Tokenizer();

I'm assuming that these tests are running just fine on your dev environment, so I don't want to go through and change all the references. Do you have any ideas about how to make these work?

from natural.

OK. I just hacked around my directory issue for now (although, I'd love to get that figured out).

from natural.

sorry for the late response and thanks for the fix.

i'll be in touch about the jasmine path issue in a few days. i more-or-less know what's going on there.

from natural.

We're good on this one, right?

from natural.

Yeah.

On Tuesday, August 7, 2012, Chris Umbel wrote:

We're good on this one, right?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/44#issuecomment-7571929.

Adam - [email protected] (new)

from natural.

sweet, thanks!

from natural.

Hi,

I would like to ask about JaroWinklerDistance algorithm.

My team made a comparison between 'jezze' and '09270000000' and it returned 0.411111111111111 value (we are expecting 0).

To be honest I am not familiar with Jaro-Winkler and it's algorithm, but upon researching about it I found out that Jaro-Winkler is really useful. My question is, maybe the implementation of the algorithm is incorrect?

Hoping for your feedback.

Thank you!

from natural.

You are correct, this should be 0.

from natural.

I tested this, and works like expected. See #491 for the same issue and my answer.

from natural.

from natural.