alternative Jaccard

[markvanderloo]

Suggested by Tom Magerman by e-mail to add

• Dice Coefficient
• Sorensen overlap

to the q-gram distances

[wynksaiddestroy]

On top of that it would be nice to be able to use the Jaccard similarity with whole words instead of q-grams. What do you think about it?

[markvanderloo]

You can tokenize.using one of the many tokenizers available in R, then hash the tokens (words) to integer using the hashr package and.then use stringdist::seq_dist.

It's basically why I wrote the hashr package, but it has gone unnoticed. I should.blog about this at some point..