Reputation: 297
Is there a way to convert a list of cosine similarities to percentage? I tried to wrap my brain around this but I'm in great doubt.
Would it make sense to normalize the cosine values of the four documents like so:
Doc #1 0.9600
Doc #2 0.9300
Doc #3 0.8800
Doc #4 0.8500
Summing them all up
0.9600 + 0.9300 + 0.8800 + 0.8500 = 3.6200
And normalize them.
Doc #1 0.9600 / 3.6200 = 0.2652
Doc #2 0.9300 / 3.6200 = 0.2570
Doc #3 0.8800 / 3.6200 = 0.2431
Doc #4 0.8500 / 3.6200 = 0.2348
or is there are more accepted way of displaying this?
Upvotes: 0
Views: 2207
Reputation: 13558
I guess it depends on your use case, but in general I don't think there's much of a need to normalize cosine similarity scores as they are already on a 0 to 1 scale.
Upvotes: 3