Comparison of local large language models for extraction of signs and symptoms data from electronic health records

Isa Spiero, Merijn H Rijk, Matthew A Scheeres, Frans H Rutten, Geert-Jan Geersing, Tamara N Platteel, Karel GM Moons, Lotty Hooft · 2026

Read Comparison of local large language models for extraction of signs and symptoms data from electronic health records by Isa Spiero, Merijn H Rijk, Matthew A Scheeres, Frans H Rutten, Geert-Jan Geersing, Tamara N Platteel, Karel GM Moons, Lotty Hooft (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0350625.txt.

Description

Isa Spiero and colleagues compared Dutch language models for pulling signs and symptoms from primary care notes. Health records hold rich detail in free text, but hand coding is slow. Two local models were trained as direct classifiers on hand labelled notes about chest infections, with training sizes varied to find how much labelled data is enough.

Direct classifiers scored well, led by the medical tuned model. Prompt style use scored poorly for both models. Rare symptoms and missing mentions dragged scores down. About eight hundred labelled examples marked the floor for solid results. Local language models, the work shows, can unlock Dutch record text for research when labelled data are sufficient.

Catalog record

CREATORIsa Spiero, Merijn H Rijk, Matthew A Scheeres, Frans H Rutten, Geert-Jan Geersing, Tamara N Platteel, Karel GM Moons, Lotty Hooft
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE33420
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0350625

Browse the full library · About the Tower of Babel