Is there a way to have NLP model predict more text instead of a category or sentiment?
Given a sentence or paragraph I’d like to remove certain tokens or change the token order based on examples of how people had ‘cleaned up’ the input by hand. Is this even possible?
Unrelated question, Is the fastai NLP api the best fit for predicting a category when there are only a few tokens (1 or 2 up to maybe 5) in the input? For example if I wanted a model that could predict that “Address” “Addr” “Location” “Customer Loc” all map to the category “Address”. It will take quiet quite some effort to create the training data, so I’m wondering if you all have any ideas before I just try it .