Right now, according to pyLDAvis chemical descriptions look nice, but there seem to be a few very big topics and some small ones ... these smaller topics are also collapsed in a short area of the 2D pyLDAvis plot ...
Is this an issue of parameter selection in sparkLDA ??
We should also revise the calculation of pyLDAvis for the neural models ...
Right now, according to pyLDAvis chemical descriptions look nice, but there seem to be a few very big topics and some small ones ... these smaller topics are also collapsed in a short area of the 2D pyLDAvis plot ...
Is this an issue of parameter selection in sparkLDA ??
We should also revise the calculation of pyLDAvis for the neural models ...