Little confusion about how tf-idf model is implemented in gensim

Question

I am working on tf-idf model. I have little confusion as how this model is implemented. I have constructed model now when I am trying to print the model it is giving different value for the same term. For following two term are giving these result:

doc_bow = [(0, 1), (1, 1)]
val1= tf_idf_corpus[doc_bow] 

doc_bow = [(0,1)]
val2=tf_idf_corpus[doc_bow] 

Following is the result:

val1= [(0, 0.56486634414605663), (1, 0.82518241210720711)]
val2=[(0, 1.0)]

I am just curious to know, why tf-idf value of term 0 is 0.5648 in val1 and 1.0 in val2.

Answer 1

The documentation may help clear your confusion: http://radimrehurek.com/gensim/models/tfidfmodel.html

I am just curious to know, why tf-idf value of term 0 is 0.5648 in val1 and 1.0 in val2.

The output vectors are normalized to unit (euclidean) length. You can turn this off, using the normalize constructor parameter.