Human Value Detection Demo

Demo for the Adam Smith human value detector by Schroter et al. (2023) [paper under review], which performed best in the ValueEval'23 competition. This demo uses the EN-Deberta-F1 ensemble of three models that performed best in the ablation tests. [code: original, docker image, server docker image]

Enter an argument in the text area and click on submit. After a few seconds, the detected value categories (and higher-level values containing them) will be highlighted in the value taxonomy wheel. See Kiesel et al. (2022) and Schwartz et al. (2012) for details on the wheel.