topicvec icon indicating copy to clipboard operation
topicvec copied to clipboard
verified publisher icon askerlee

LogLikelihood

Open gabrer opened this issue 8 years ago • 2 comments

Hi askerlee!

I would ask you if the logLikelihood computed by "calcLoglikelihood" function is normalised by the number of words in the corpus? If not, it could be easily done?

Thank you!

gabrer avatar Jun 06 '17 15:06 gabrer

It's not normalized by the document length (i.e. the number of words in a document). You could divide it by the document length to get the average (log) perplexity.

askerlee avatar Jun 07 '17 04:06 askerlee

So, I need only to divide it by the document length, great!

Thank you for your prompt reply!

gabrer avatar Jun 08 '17 12:06 gabrer