Combining contextualized word representation and sub-document level analysis through Bi-LSTM+CRF architecture for clinical de-identification | Publicación