As said in the post, the machine learning algorithm was trained a small set, the main purpose of showing this is to receive user feedback to make sure these kind of errors do not happen. I think I made it clear that for now the algorithm is far from perfect.