Fig. 1
From: De-identifying free text of Japanese electronic health records
Conceptual figure of our LSTM-based model, showing embedding and NER in separate figures. + means concatenation. The first figure shows the embedding part, where Wx is an xth input word, Lx,i is an ith letter of the word Wx, r denotes right to left (forward) LSTM, l denotes left to right (backward) LSTM, Vx is an intermediate node which corresponds to Wx. The second figure shows the NER part, where fl denotes forward LSTM, bl denotes backward LSTM, c denotes concatenated vector, finally a CRF layer is shown with an example predicted named entities in the BIO annotation style