Developing a natural language processing system using transformer-based models for adverse drug event detection in electronic health records
Jingyuan Wu, Xiaodi Ruan, Elizabeth McNeer, Katelyn M Rossow, Leena Choi · 2026
Read Developing a natural language processing system using transformer-based models for adverse drug event detection in electronic health records by Jingyuan Wu, Xiaodi Ruan, Elizabeth McNeer, Katelyn M Rossow, Leena Choi (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-0350516.txt.
Description
Jingyuan Wu and four co-authors present this 2026 PLOS One study teaching language models to spot harmful drug reactions. The clues hide in doctors' free-text notes. The team worked from a national benchmark dataset and a second set from Vanderbilt University Medical Center. They fine-tuned several transformer models. They compared two ways of chopping long notes into pieces the models can digest.
Cutting notes into structured chunks generally beat sliding a fixed word window across them. One long-document model, paired with chunking, gave the most balanced results across both datasets. Chunk size mattered even when notes fit within model limits. Scores fell on the real hospital data compared with the benchmark. The authors report that gap honestly, as they offer practical guidance for building such safety tools.
Catalog record
| CREATOR | Jingyuan Wu, Xiaodi Ruan, Elizabeth McNeer, Katelyn M Rossow, Leena Choi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36505 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |