How to normalize LCS algorithm result in a way that strings's length won't affect?

Asked Jun 10 '16 at 17:15

Active Jun 12 '16 at 23:32

Viewed 336 times

I have n strings (each has its own size) and the letters are included in finite group - S (~ 120 letters).
I want to calculate LCS result between each string to another and I want that all results will be normalized.

I want to normalize LCS result between string i and string j to avoid both strings's length.

Example:
LCS("shpin","shdek")=2 because ("[sh]pin","[sh]dek") = "sh"
but
LCS("shpxaaaaaaaaaan","shaaaaaaaaaadek")=12 because ("[sh]px[aaaaaaaaaa]n","[shaaaaaaaaaa]dek") = "shaaaaaaaaaa"

I was thinking about dividing each result in Expected Value but I don't know to calculate EV.
Does anyone have a solution? maybe another way to get good enough approximation? :(

Thanks

edited Jun 12 '16 at 23:32

asked Jun 10 '16 at 17:15

user3781974

Could you maybe include sample input and expected output? – Martin Gottweis Jun 10 '16 at 17:23
What do you mean by "normalization?" – templatetypedef Jun 10 '16 at 18:21
I got a mistake, I don't know how to calculate Expected Value between two random sequences. does anyone know? maybe another way to get a good approximate result? :( – user3781974 Jun 12 '16 at 21:35

How to normalize LCS algorithm result in a way that strings's length won't affect?

0 Answers0