Lda2vec: Tools for interpreting natural language

Recent: https://news.ycombinator.com/item?id=10919744

Very cool indeed. I wonder if it permits incremental model generation, so that i could be usable for classifying new data and new topics?

How does lda2vec compare with sense2vec [1]?

[1] http://arxiv.org/abs/1511.06388

sense2vec tags tokens with information from syntactic parsing and by that offers finer grained meaning on the word level. Here is a good implementation btw: https://spacy.io/blog/sense2vec-with-spacy

lda2vec has a different goal, namely classifying utterances with features on the sentence/document level. Thus it is to my understanding doc2vec with regularization for enhanced interpretability.