Currently, the model handles multi-context lda2vec. That's cool, but it makes the regular version extremely hard to read/understand. Making 2 versions and/or subclassing the original model would allow for the regular version to be more understandable to new users.
Currently, the model handles multi-context lda2vec. That's cool, but it makes the regular version extremely hard to read/understand. Making 2 versions and/or subclassing the original model would allow for the regular version to be more understandable to new users.