Leveraging Open-Source Large Language Models for Data Augmentation in Hospital Staff Surveys: Mixed Methods Study
Carl Ehrett, Sudeep Hegde, Kwame Andre, Dixizi Liu, Timothy Wilson, Blake Lesselroth · 2024
Read Leveraging Open-Source Large Language Models for Data Augmentation in Hospital Staff Surveys: Mixed Methods Study by Carl Ehrett, Sudeep Hegde, Kwame Andre, Dixizi Liu, Timothy Wilson, Blake Lesselroth (2024) 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/pmc-11590755.txt.
Description
Small survey datasets are hard to classify well, and hospitals cannot always ship staff comments to commercial AI services. Carl Ehrett and co-authors tested in 2024 whether open-source language models could manufacture extra training text. Their material was survey narratives from frontline radiology staff describing how they adapted during the first COVID-19 wave.
Four generative models produced synthetic responses in the same style; three separate classifiers then sorted real and synthetic text into themes. The best recipe paired a LLaMA 7B generator, set to a middling randomness, with one hundred synthetic cases and a RoBERTa classifier, reaching an average accuracy score of 0.87. The authors conclude that open models can lift classifier performance on small health datasets while keeping data in-house. They also flag privacy and ethics questions for further work.
Catalog record
| CREATOR | Carl Ehrett, Sudeep Hegde, Kwame Andre, Dixizi Liu, Timothy Wilson, Blake Lesselroth |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |