--------------- For the penniless uneducated machine learning engineer, a good start here would be producing a classifier for whether expressions are mediative or not.
An example work for doing this very simply could be my "adventure3.py" uhh under the thread "middle smart adventure". It uses prompts to generate classifiers, optionally based on examples. [examples are chosen from tasks the system otherwise fails at]. That's a prototype approach, and then one uses (prompt tuning, embedding tuning, model adapters, active learning, model finetuning, model training) to produce a stronger solution once the prototype is put together. The other file in "middle smart adventure" uhhh the one that generates stuff, shows an approach that could be used for creatively finding solutions to hard situations.
