#pretrained-embeddings

Question 1: Select all that apply

You must choose between word2vec, fastText and pretrained embeddings for a text project. Which statements are true? Select all that apply.

With a vocabulary of 20,000 words, each word2vec embedding has 20,000 dimensions to capture the word's full meaning1
A word2vec model has no vector for a word that was not in its training vocabulary2
A fastText model can give a vector even for a word that was not in its training vocabulary3
For a huge biomedical corpus full of domain terms and abbreviations, training your own embeddings beats using Google News embeddings4
If new text keeps bringing unknown words, fastText (trained yourself or pretrained) is a sensible choice5