I have a huge doubts about text pre-preparation stage using deep learning.
I understand that neurals networks are able to take the raw data, extract features and make a prediction or classification, we don’t need to extract the features “manually”
I know, that the input does not need to be pre-prepared (i.e. : removing stop words, stemming / lemmatization, tokenization, word embedding and other related techniques belonging to regular machine learning or past techniques)
I see that applying this pre-preration steps, I am trying to help the neurals networks and with this intention I could cause a bad accuracy
Is it correct ?
Thanks a lot