Research map / Label-efficient and robust learning / Recognition and long-tail
Supervised Word Mover's Distance
Gao Huang, Chuan Guo, Matt J Kusner, Yu Sun, Fei Sha, Kilian Q. Weinberger
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 |
|---|---|---|---|
| supervised training of metric and word weights (S-WMD vs S-WMD init.) (ablation) | S-WMD · OHSUMED | kNN error: 38.0 -> 34.3 | Improved |
| supervised training (S-WMD vs S-WMD init.) (ablation) | S-WMD · BBCSPORT | kNN error: 2.8 -> 2.1 | Improved |
| supervised training (S-WMD vs S-WMD init.) (ablation) | S-WMD · REUTERS | kNN error: 3.5 -> 3.2 | Improved |
| supervised training (S-WMD vs S-WMD init.) (ablation) | S-WMD · AMAZON | kNN error: 5.8 -> 5.8 | No clear effect |
| supervised training (S-WMD vs S-WMD init.) (ablation) | S-WMD · 20NEWS | kNN error: 28.4 -> 26.8 | Improved |
| supervised WMD vs unsupervised WMD | S-WMD · TWITTER | kNN error: 28.7 -> 27.5 | Improved |
| supervised WMD vs unsupervised WMD | S-WMD · RECIPE | kNN error: 42.6 -> 39.2 | Improved |
| supervised WMD vs unsupervised WMD | S-WMD · CLASSIC | kNN error: 2.8 -> 3.2 | Hurt |
| S-WMD vs LSI on datasets where LSI beat WMD | S-WMD · BBCSPORT, OHSUMED | kNN error: relative reduction of 51% and 22% | Improved |
| S-WMD vs NCA with TF-IDF on short-document datasets | S-WMD · CLASSIC | kNN error: 5.5 -> 3.2 | Improved |
| S-WMD vs NCA with TF-IDF on long-document datasets | S-WMD · BBCSPORT | kNN error: 0.6 -> 2.1 | Hurt |
| S-WMD vs NCA with TF-IDF | S-WMD · OHSUMED | kNN error: 35.8 -> 34.3 | Improved |
Closest papers by meaning
- Re-evaluating Word Mover’s Distance (ICML 2022 · similarity 0.91)
- Hierarchical Optimal Transport for Document Representation (NeurIPS 2019 · similarity 0.88)
- UltraTWD: Optimizing Ultrametric Trees for Tree-Wasserstein Distance (ICML 2025 · similarity 0.83)
- Contextual Document Embeddings (ICLR 2025 · similarity 0.82)
- Analyzing Similarity Metrics for Data Selection for Language Model Pretraining (NeurIPS 2025 · similarity 0.82)
- Repurposing Language Models into Embedding Models: Finding the Compute-Optimal Recipe (NeurIPS 2024 · 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.
