I see! Thanks for the feedback... expanding the training data by reasonable but not too biased examples will be the major challenge (as it seems to be with AI).

I'm curious to see how the detection will improve over time. Thanks a ton for the efforts you are making on this!