This will contain collectively generated notes
The forloop presented in the pseudo code used to compute the cosine similarity between two documents (using tf-idf) is over words, which is the faster way. If the outer level of the forloop were over documents, it would be slower.
-Jadiel
