Fig. 6From: Multitask learning for biomedical named entity recognition with cross-sharing structureCross-sharing Multi-task Model (CS-MTM). The embedding layer and shared Bi-LSTM are shared by two datasets. Gated interaction unit is used to adjust the output of private Bi-LSTMs. P1,P2: Output of private Bi-LSTMs. S: Output of the shared Bi-LSTM. G1,G2: Output of the gated interaction unitBack to article page