Research map / Label-efficient and robust learning / Noisy labels
IWE-Net: Instance Weight Network for Locating Negative Comments and its application to improve Traffic User Experience
Lan-Zhe Guo, Feng Kuang, Zhang-Xun Liu, Yu-Feng Li, Nan Ma, Xiao-Hu Qie
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use IWE-Net instead of XGBoost | XGBoost · Didi negative comment test set | AUC (%): 79.55 -> 91.24 | Improved |
| use IWE-Net instead of LTR | LTR · Didi negative comment test set | AUC (%): 85.45 -> 91.24 | Improved |
| use IWE-Net instead of GLC | GLC · Didi negative comment test set | AUC (%): 82.63 -> 91.24 | Improved |
| use IWE-Net instead of Rank Pruning | Rank Pruning · Didi negative comment test set | AUC (%): 82.22 -> 91.24 | Improved |
| train XGBoost on validation data only instead of train plus validation (ablation) | XGBoost · Didi negative comment test set | AUC (%): 79.55 -> 60.98 | Hurt |
| apply label bias correction to label noise learning methods (ablation) | Rank Pruning, GLC, LTR · Didi negative comment test set | AUC (%) | No clear effect |
| vary positive to negative ratio in validation set from 1:1 to over 40:1 (ablation) | IWE-Net vs LTR and GLC · Didi negative comment test set | AUC (%) | Improved |
| use IWE-Net instead of LTR on running time | LTR · Didi negative comment test set | time cost (s): 211.6 -> 107.1 | Improved |
| use IWE-Net instead of Rank Pruning on running time | Rank Pruning · Didi negative comment test set | time cost (s): 601.4 -> 107.1 | Improved |
| use IWE-Net instead of DNN on running time | DNN · Didi negative comment test set | time cost (s): 90.4 -> 107.1 | Mixed |
| use IWE-Net instead of LTR on F1 | LTR · Didi negative comment test set | F1-score: 15.70% -> 27.41% | Improved |
| use IWE-Net instead of GLC on precision | GLC · Didi negative comment test set | precision: 24.13% -> 16.50% | Hurt |
Closest papers by meaning
- Enhancing Learning with Noisy Labels via Rockafellian Relaxation (ICLR 2026 · similarity 0.82)
- Interactive Weak Supervision: Learning Useful Heuristics for Data Labeling (ICLR 2021 · similarity 0.82)
- Weak Supervision Performance Evaluation via Partial Identification (NeurIPS 2024 · similarity 0.82)
- IWBVT: Instance Weighting-based Bias-Variance Trade-off for Crowdsourcing (NeurIPS 2024 · similarity 0.82)
- Learning from weak labelers as constraints (ICLR 2025 · similarity 0.82)
- Ordinal Programmatic Weak Supervision and Crowdsourcing for Estimating Cognitive States (Student Abstract) (AAAI 2023 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
