> they seem to have dipped a toe into statistical language modelling but be unwilling to go all the way with more advanced approaches

That's fair.

Thanks for your suggestions. I clearly have some studying to do.