For text classification with machine learning - If your training data was generated purely with regular expressions, is it possible to train a machine learning model with this training data which will be more accurate than your original regex?
- By 'accuracy' I mean to say the percentage that are correctly labelled as compared to the human truth of the question.
(E.g. we have a set of Tweets that we've labelled as happy or sad with regex. I'm wondering if a machine learning classifier trained on this data will be able to surpass the accuracy of the regex that generated the training data)