My org (a university) is investigating creating a chatbot, first to handle common questions but utlimately to serve as a ML-driven tutor. We've had a number of conversations with solution architects at Amazon who've pointed us to Lex, but I'm either missing something or Lex isn't well suited for the level of context-awareness we want without significant custom coding, and the other bot frameworks seem about the same. Can I make a chat bot from scratch without being Amazon, Google, or Facebook? Is anyone using Clojure's nlp and deep learning libraries to do this already?
Figure I will post answer here as well: See Flare https://github.com/aria42/flare which is focused on sequence analysis with LSTM capabilities on top of Neanderthal. Similar to PyTorch in interaction and use. Actually, very good performance (since it is on Neanderthal which uses MKL/CUDA/OpenCL). @d4hines Even though it is not as close a 'subject' match, #data-science is probably a 'better' general purpose place to post for this stuff - overall more activity.