Weird behavior about string-similarity

[aceakash]

That's how the algorithm works -
NOS is split into two bigrams NO and OS
NPS is split similarly into NP and PS
There are no common bigrams, hence similarity is 0

Further reading: https://en.wikipedia.org/wiki/S%C3%B8rensen%E2%80%93Dice_coefficient#Formula

[jfmmm]

Ah I guess I should have read more on the algoritm.

I ended up using this instead and it did what I wanted.
https://www.npmjs.com/package/leven

Thanks for the answer.