Vectorize only relevant documents per query

[krinkv]

Using inverse index we can obtain documents that contain at least one word from the query

Note: Later we can think about vectorizing the document on initialization