Fig. 3
From: Automatic literature screening using the PAJO deep-learning model for clinical practice guidelines
PAJO network architecture. Each article’s raw title and abstract are fed into the PubMedBERT text encoder for conversion to embedding vectors. The vectors are passed to an attention encoder for weight sample representation. The original journal features are normalized and passed to a feed-forward layer with a rectified linear unit (ReLU) activation function. The obtained text and journal features are concatenated to obtain the overall feature representation of an article. Finally, the feature representation is passed to a feed-forward layer with a SoftMax function to predict the article’s label