DARE UK Next-Gen Catalyst

TRExt: TRE Text Analytics

Bringing natural language processing and data standardisation capabilities to Trusted Research Environments to make sensitive unstructured text federation-ready.

Explore the Research DARE UK Project Page ↗

About the Project

Project Overview

Sensitive free-text holds vital research evidence that current Trusted Research Environments (TREs) cannot access. TRExt unlocks it by converting unstructured documents into structured, standardised data that can be used in federated analyses. Sensitive data never leaves the TRE. We demonstrate this capability using clinical text.

The Challenge

Some of the richest health information available is recorded in unstructured text, including clinical letters, discharge summaries, and radiology reports. Yet converting this free-text into the standard data models used by TREs, at the scale demanded by modern research, remains an unsolved technical challenge. As a result, unstructured clinical text is rarely included in TREs, and the evidence it contains remains out of reach for researchers.

Our Approach

TRExt embeds a modular natural language processing pipeline directly inside a TRE, automatically converting sensitive clinical free-text into structured, standardised data conforming to a data model used by TREs (called OMOP). The pipeline provides Safe Data for use with federated data technology (called Five Safes TES), enabling the resulting structured data to be queried across institutions as part of federated analyses.

Public Involvement & Engagement

Involving the public is central to TRExt. We run a structured programme of workshops and co-design activities aligned to DARE UK public involvement and engagement guidelines, ensuring that public perspectives shape technical design decisions throughout the project.

Expected Impact

TRExt will unlock federated analysis of unstructured senstive text for the first time within TREs, opening new research possibilities across a wide range of clinical domains. Beyond its immediate outputs, the project will establish reusable, open-source infrastructure and set standards for responsible, privacy-preserving text analytics that any TRE in the UK can adopt.