Збережено в:
| Автор: | |
|---|---|
| Формат: | Recurso digital |
| Мова: | Англійська |
| Опубліковано: |
Zenodo
2026
|
| Предмети: | |
| Онлайн доступ: | https://doi.org/10.5281/zenodo.19717749 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
Зміст:
- <p>NLP pipeline for detecting fraud-associated characteristics (tortured phrases, formulaic structure, AI-generated text markers, citation anomalies, co-authorship network patterns, cross-document similarity) in the medical artificial intelligence literature. Trains a supervised classifier using Retraction Watch labels and estimates prevalence via Positive-Unlabelled learning correction. v3.0.0 adds a leave-Hindawi-out sensitivity analysis script; no changes to pre-registered classifier, weights, or feature definitions. v2.0.0 includes bug fixes identified during analysis (see CHANGELOG.md). Analysis plan pre-registered on OSF (DOI: 10.17605/OSF.IO/JB4T6). v1.0.0 (pre-registered code): DOI 10.5281/zenodo.19481250.</p>