Fig. 2From: Multitask learning for biomedical named entity recognition with cross-sharing structureSingle-task Model (STM). The input is a sentence from the BioNER dataset. The dotted rectangles represent words in a sentence, and the solid rectangles represent Bi-LSTM cells. The circles represent CNN units, and the double circles represent CRF units. The tags in the double circles, e.g., “O”, “B-GENE”, are the output of the CRF layerBack to article page