How tf-idf is relevant in calculating sentence vectors

Question

I am interested to find sentence vectors using word vectors.I read that by multiplying each word's tf-idf weights with their vectors and finding their average we can get whole sentence vector.

Now I want to know that how these tf-idf weights helps us to get sentence vectors ie how these tf-idf and sentence vector are related?

Answer 1

Any aggregative operation on the word vectors can give you a sentence vector. You should consider what do you want your representation to mean and choose the operation accordingly.

Possible operations are summing the vectors, averaging them, concatenating, etc.