LSA - steps after finding the SVD

Question

I have read quite a few tutorials since morning . My problem involves finding the similarity between two documents. I am looking forward to use LSA in java for this purpose.

I understood the creation of the term-document matrix and then the SVD(Dimensionality gets reduced) is applied to it . 3 Matrices are obtained as results.This might sound stupid but i have been stuck with this for a quite a while . Now if i have to find the similarity between the two documents what do i have to do ?

since i am comparing only two documents will it be advantageous to use LSA ? — CTsiddharth, Jan 23 '12 at 03:13

score 1 · Answer 1 · answered Jan 28 '12 at 02:02

1

After calculating the 3 matrices using SVD, you need to calculate the correlation between the vectors of the two documents you want to compare. you can use spearman's correlation. Another way is with using the cosine distance.

you will find more details at LSA, there is a full example with explanation.

you might search for some java libraries for LSA.

answered Jan 28 '12 at 02:02

Samer Makary

1,815
2
22
25

Hi i recently came across this link . http://www.miislita.com/information-retrieval-tutorial/svd-lsi-tutorial-4-lsi-how-to-calculations.html . this approach seems to be simple and serves my purpose perfectly – CTsiddharth Mar 07 '12 at 10:48

LSA - steps after finding the SVD

1 Answers1