One possible trick could be to search and replace them all with nonsense alternatives then see if it extracts those.
That might actually boost performance since attention pays attention to stuff that stands out. If I make a typo, the models often hyperfixate on it.
That might actually boost performance since attention pays attention to stuff that stands out. If I make a typo, the models often hyperfixate on it.