Figure 2 | Scientific Reports

Figure 2

From: Knowledge grounded medical dialogue generation using augmented graphs

Figure 2

MED architecture: The proposed architecture first encodes the conversation history using a BioBERT encoder as shown in the left. We first attempt to predict the sequence \(M_L\) using the token classification task that provides the hidden states, \(H_{k,i}\), for the input sequence. This input sequence is then fed into a BioBERT decoder to generate the response.

Back to article page