Reproduce “AutoMeTS: The Autocomplete for Medical Text Simplification” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hoang Van, David Kauchak, Gondy Leroy
Researchers from the University of Arizona and Pomona College introduced AutoMeTS, an autocomplete system designed to assist human experts in simplifying medical texts. AutoMeTS combines fine-tuned pretrained neural language models with a multi-label classifier ensemble, achieving 64.52% accuracy in predicting the next word, and also contributes a new 3.3K sentence pair medical parallel corpus.
Key insights
- Incorporating the original, difficult source sentence as additional context significantly improves the accuracy of PNLM-based autocomplete suggestions for text simplification.
- Modern PNLMs can be effectively adapted for interactive, semi-automated text simplification, with RoBERTa showing the strongest individual performance (62.40% accuracy with context).
- Ensemble methods, particularly the proposed AutoMeTS multi-label classification approach, can outperform the best individual PNLMs by leveraging diverse model predictions and mitigating selection biases.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.