APEX-VW: A Document-Level English-Spanish Post-Editing Dataset in the Healthcare Domain

APEX-VW (Automatic Post-Editing eXperiments on Virtual Wards) is a document-level English-Spanish (EN–ES) Post-Editing (PE) dataset designed to support research on correction propagation, terminology normalisation, and consistency support in Machine Translation (MT) workflows in the healthcare domain.

The corpus is built from seven openly licensed NHS and related UK public-sector documents about virtual wards and closely related service models. These texts (≈42k English source words) cover topics such as hospital-at-home pathways, diagnostics, pharmacy services, and technology-enabled virtual wards.

All sources were cleaned and segmented for professional translation in Trados Studio. Each document was translated with one of four MT systems (DeepL, ModernMT, Language Weaver, and an OpenAI-based GPT-5 system), and then fully post-edited by professional EN-ES translators under controlled conditions.  All MT outputs were generated on 18 April 2026, using the then-current production models exposed through each provider's integrations in Trados Studio. 

The dataset includes:

• The segmented corpus in CSV format with 2,712 rows. Each row corresponds to one segment and contains:

  •  globalid: id assigned to the segment at the dataset level

  • docsegmentno: segment number at the document level

  • tradossegmentid: segment id in Trados Studio

  • source: source segment (English)

  • targetmt: raw MT output (Spanish)

  • targetpe: post-edited version (Spanish)

  • sourcetextid: id assigned to the source text

  • mtsystem: system used to generarte the MT output

• A terminology CSV file with 215 term entries extracted and consolidated during post-editing.

• An acronym CSV file with 79 items covering the main domain-specific abbreviations used in the corpus.

• Documentation files

  • README

  • Datasheet

APEX-VW is intended for researchers and practitioners working on Automatic Post-Editing, Quality Estimation, Terminology and Consistency modelling, document-level Machine Translation, and human-in-the-loop translation tools, especially in specialised and safety-critical domains such as healthcare.