GRAPHTREX is a novel span-based graph transformer that significantly improves extraction of temporal relationships in clinical documents, enabling better contextual understanding of medical events.

• Achieves state-of-the-art performance on the I2B2 2012 Temporal Relations Challenge
• Effectively handles complex clinical language and long documents with sparse temporal annotations
• Integrates span-based entity-relation extraction with graph transformer architecture
• Enables more accurate sequencing of clinical events for improved diagnostic and prognostic modeling

This research provides critical advancement for clinical decision support systems by automatically extracting when medical events occurred relative to each other—essential for proper treatment planning and medical record interpretation.

Temporal Relation Extraction in Clinical Texts: A Span-based Graph Transformer Approach