Which of the following operations can be skipped in ranked retrieval when using the vector space model?
1) Computing cosine similarity for top documents.
2) Selecting/sorting top scores.
3) Normalizing query vectors.
4) Normalizing document vectors.