I have figured out how to explain the simple version of such a model using histograms, even able to explain the concept of embeddings, but I am stuck on the final step - how to explain and visualise the modelling process.
My flow shows that histograms using bi-grams, tri-grams, n-grams cannot scale - we need too many variables. So a model is used, but I am not able to explain this final leap in a non-technical manner.