About DeepVigilace

Drug safety is the main goal of pharmacovigilance, yet the field is plagued by the constraints of textual information processing. In this project, we adapted contrastive learning methods to create vector representations, i.e. embeddings, of adverse events, and trained a deep neural network classifier to determine the causal relation of drug–event pairs.

The aim of the project is to help further the cause of computer-assisted causality assessment for authorities and field experts in pharmacovigilance. Therefore, we provide open access to our framework as well as adverse event embeddings, ready for machine learning applications in research and development.