Core Concepts
Incorporating medical knowledge into graph neural networks improves ADE detection performance.
Abstract
Adverse drug events (ADEs) are crucial for drug safety monitoring.
Text-based automated ADE detection is essential due to limitations in clinical trials.
Recent studies focus on using text data from various sources for ADE detection.
Knowledge-augmented graph neural networks with concept-aware attention improve ADE detection.
Different computational methods impact edge weights in graph construction.
Various graph architectures perform differently based on dataset characteristics.
Concept-aware attention consistently enhances model performance across datasets.
Stats
"Experiments on four public datasets show that our model performs competitively to recent advances."
"The concept-aware attention consistently outperforms other attention mechanisms."