PURE

Why we included this project

PURE is a PyTorch implementation of a well-known approach to entity and relation extraction that keeps the two tasks separate. The entity model tags everything in a sentence at once, and the relation model then inserts typed markers around each entity pair and predicts the relation between them, with a batched approximation variant for faster inference. Pre-trained models for ACE04, ACE05, and SciERC are included, so you can run the paper's evaluation setup without training from scratch. The code is easy to follow and the data formats are clearly documented, which makes it a practical starting point if you want to adapt the method to your own extraction pipeline.