Temporal ordering of clinical events

TL;DR

This paper presents a set of minimalistic methods for extracting and temporally anchoring clinical events, expressions, and links from medical texts, validated on high-quality datasets.

Contribution

It introduces new methods to identify and link clinical events and temporal information, improving temporal understanding in clinical narratives.

Findings

Effective extraction of clinical events and temporal expressions

Accurate temporal linking between events and expressions

Validated methods on high-quality datasets

Abstract

This report describes a minimalistic set of methods engineered to anchor clinical events onto a temporal space. Specifically, we describe methods to extract clinical events (e.g., Problems, Treatments and Tests), temporal expressions (i.e., time, date, duration, and frequency), and temporal links (e.g., Before, After, Overlap) between events and temporal entities. These methods are developed and validated using high quality datasets.