Possible to do exploratory factor analysis on word embeddings?

Question

I am trying to develop linguistic scales to measure the extent of \'market\' vs. \'social\' language used in a text. Here are the two proposed scales:

s1= [\'mark