The text explains that uncertainty is a fundamental part of scientific writing, yet difficult to detect automatically. The new system “UnScientify”, developed by researchers from France and Germany, uses transparent, rule-based methods to identify and categorize different forms of scientific uncertainty. It even distinguishes whether the uncertainty is expressed by the authors themselves or by others. In benchmark tests, UnScientify outperformed advanced AI models like GPT-4 in both accuracy and explainability. The tool aims to help science and society better identify uncertainty within research texts, promoting greater transparency and more effective knowledge transfer.
New on the GESIS Blog: The new post shows how “UnScientify” can help to better identify uncertainties in research texts
Categories:
GESIS-News
