[Corpora-List] LSA and number of factors

[Matthew T. Bell]

Dear Corpora list,

I'm performing an experiment with LSA, and notice that the scripts
provided in the LSA distribution allow independent tweaking of the number
of factors used in the indexing and the querying steps, i.e. there is a
script for performing indexing (pindex), and another for performing
querying (Q), and both allow me to set the number of factors considered.

Does it make any sense, given these are independent, to vary the number of
factors at both steps independently for parameter testing/tuning?

Thanks in advance for any advice!

Regards,

Matthew Bell
Graduate Student
University of Pittsburgh