Figure 3.

Architecture of the CNN-RNN model. “w” and “z” represent word embedding and POS embedding, respectively. “d” and “a” represent position embeddings. “hc,” “hr,” “lc,” “lr,” and “o” are outputs of layers in the architecture.