You are working on a text analytics project and need a way to represent words and documents numerically so similar language patterns are captured better than with simple bag of words features. You are considering word embeddings as the core text representation for downstream analysis.
How would you use word embeddings in a text analytics project?