I am trying to build an embedding for a corpus using Python's gensim's word2vec implementation. The catch is that I wish to have in the same embedding all of the unigrams and bigrams for the corpus. Is there a way to embed in the same space both unigrams and bigrams?
1 Answers
You can do it with Phrases model in gensim
from gensim.models.phrases import Phrases, Phraser
#documents is list is list of tokens from your text
bigram = Phrases(documents, min_count=2)
trigram = Phrases(bigram[documents], min_count=1)
Related