notesum.ai
Published at November 4DeMod: A Holistic Tool with Explainable Detection and Personalized Modification for Toxicity Censorship
cs.HC
cs.AI
cs.SI
Released Date: November 4, 2024
Authors: Yaqiong Li1, Peng Zhang1, Hansu Gu2, Tun Lu3, Siyuan Qiao1, Yubo Shu1, Yiyang Shao1, Ning Gu1
Aff.: 1Fudan University, China; 2Independent, USA; 3Shanghai Key Laboratory of Data Science, China and Fudan University, China

| Design goal | Description | Metric | Method | Dimension |
| G1 | Provide holistic censorship | Holistic integrity | H | Function |
| G2 | Offer fine-grained detection results | Granularity integrity | H | Function |
| Detection accuracy | A & H | Performance | ||
| G3 | Strengthen interpretability | Explanation integrity | H | Function |
| Explanation accuracy | H | Performance | ||
| G4 | Give personalized revising suggestions | Modification integrity | H | Function |
| Modification Accuracy | A & H | Performance | ||
| Semantic retention | H | Performance | ||
| Personalized degree | H | Performance | ||
| G5 | Ensure user-control | Ease of use | H | Design |
| Controllability | H | Design |