Open smolendawid opened 6 years ago
I have the same doubt. Did you have any progress about this theme?
Same doubt here. Did any of you manage to get this working?
I'm using GloVe pre-trained Twitter word embedding together with this lda2vec. Just replace the bin file here in twenty_newsgroups with the pre-trained result file. It can be txt instead of bin, the first line being the total # word entries and the dimension of the representation, separated with a space char.
Is it possible to initialize lda2vec with some external embeddings, f.e pre-trained FastText or pre-trained word2vec?