0
$\begingroup$

Thanks for stopping by. I have a directional question - I've built a Latent Dirichlet Allocation using Gensims Mallet wrapper. I trained the model once on OldDataSet.csv and measured coherence. I have been using it to pass NewDataSet.csv through for topic allocation. I need some guidance on how I might be able to predict how accurately my pre-trained model is allocating NewDataSet.csv. That coherence score only checks the accuracy of the pre-trained model not the allocated data set. I'd like a way to track the occurrence of historical topics and detect the emergence of new topics without re-training the model. Like say these are the topics in OldDataSet.csv:

  1. whiskey
  2. Tango
  3. Foxtrot

It will assign NewDataSet.csv 1. whiskey 2. Tango or 3. Foxtrot but a more accurate allocation might be:

  1. whiskey
  2. Tango
  3. Alpha

If I keep running the same model I might miss this new topic. If there exists a numeric score that would measure how closely the topics adhere to NewDataSet.csv that would be a huge time saver. Thanks Stack you always save me :)

$\endgroup$
0
$\begingroup$

I found a solution it is called dynamic topic modeling. I've linked an article documenting its' use. It is still undergoing research, but it's basically an LDA that takes time into account and can print topics change over time.

https://github.com/rare-technologies/gensim/blob/develop/docs/notebooks/ldaseqmodel.ipynb

Also check out Bleis' google talk on the matter:

https://www.youtube.com/watch?v=7BMsuyBPx90

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.