try text classifying with embeddings in the recipe

[milandv]

from Slack: https://p3lab.slack.com/archives/G014F336K5G/p1620245318208600

[jaeyk]

@milandv I added a new feature for the apply_basic_recipe() function. If you add add_embedding = TRUE, the function uses word embedding (Stanford NLP glove) in place of TF-IDF for feature engineering. I haven't tested the performance between these two options yet. My hunch is it will vary by context (as always!).