A2:cosine score() in score.py

[changediyasunny]

Prof,

Cosine sim(query,doc) = query idf weight * document tf-idf weight/ (query norm) + (doc-norm)

We have norm in index.computer_doc_norm, but what about query norms???
Also norms =[ ] with starting index 0. And doc_id starts at index 1. What is use of norm[0] value?
Because we will find norms according to norm[doc_id].

Very much confused?

[aronwc]

As we discussed, there is no need to compute the query norm, since it does not affect rankings.

What is use of norm[0] value?

In that doctest, the index contains a document with id 0

>>> norms = Index().compute_doc_norms({'a': [[0, 3]], 'b': [[0, 4], [1, 5]]}, 2, {'a': 1, 'b': 2})

Here, document 0 has values {'a': 3, 'b': 4}.

Thus, norms[0] is the norm for the document with id 0. (Recall that the return value is a dict from document id to norm). When running on the real dataset, there will be no document with id 0, since we start at 1.

[changediyasunny]

Prof,
In Comment section, it shows norm[0] = 0.444... so I thought it must be a list returning tf-weights. Also You mentioned that doc_id should start from 1 and not 0. So we considered it from 1 not paying attention that inside of TIME file, we have *FILE 1 as a starting line