Skip to main content
Fig. 1 | BMC Medical Informatics and Decision Making

Fig. 1

From: AutoDiscern: rating the quality of online health information with hierarchical encoder attention-based neural networks

Fig. 1

Overview of the HEA neural network architecture. In lieu of traditional feature engineering, the HEA architecture learns representations at the word, sentence, and document level before making a classification. Word representations are generated by the pre-trained BERT embedder. An attention mechanism aids in learning a document representation from amongst many sentences

Back to article page