Score computation improvement

[minottic]

I think score computation can be simplified.

From what I understand, you take the TFIDF(I,t) matrix and do the product with a matrix of 1 column of 1s.

If this is correct, the cosine similarity of one item should be the same as summing all the non-zero elements of this document in the TFIDF and dividing them by (the sum of their square * number of terms)

[nitrosx]

@minottic I have to admit that I do not remember the details of the cosine similarity method (although I studied them when I coded the scoring system).
If your suggestion holds, I will be happy to implement your solution.
Do you have any suggestion on how we can test it?
Thanks