Combination of Deep Recurrent Neural Networks and Conditional Random Fields for Extracting Adverse Drug Reactions from User Reviews

<div>The main architecture of our model. Word embeddings are given as input to the bidirectional LSTM network. Dashed arrows represent the input and output vectors of the network with dropout. The labels follow the BIO (Beginning Inside Outside) tagging scheme.</div>

Journal of Healthcare Engineering

fig1

Figure 1

Figure 1: Combination of Deep Recurrent Neural Networks and Conditional Random Fields for Extracting Adverse Drug Reactions from User Reviews 