First I would like to thank you for this awesome library, it's handy and very useful! I wanted to calculate some evaluation metrics (like perplexity, topic similarity or the log likelihood on held-out data), but unfortunately I could only find the log likelihood metric implemented.
Is there the possibility to calculate some more evaluation metrics? Am I missing anything? And if not, is there the possibility to calculate them by myself using the existing parameters and functions of the model? I'm quite new to topic modelling, so maybe it is easily possible and I'm just missing the point ... :-)!
I have no immediate solution for this Python module (which apparently has a lot of installation issues anyway). But, you could check what we are doing with the R version called seededLDAhere.
Hi guys!
First I would like to thank you for this awesome library, it's handy and very useful! I wanted to calculate some evaluation metrics (like perplexity, topic similarity or the log likelihood on held-out data), but unfortunately I could only find the log likelihood metric implemented.
Is there the possibility to calculate some more evaluation metrics? Am I missing anything? And if not, is there the possibility to calculate them by myself using the existing parameters and functions of the model? I'm quite new to topic modelling, so maybe it is easily possible and I'm just missing the point ... :-)!
Would love to here from you guys! Cheers