Research map / Label-efficient and robust learning / Recognition and long-tail
All-in Text: Learning Document, Label, and Word Representations Jointly
Jinseok Nam, Eneldo Loza Mencía, Johannes Fürnkranz
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 |
|---|---|---|---|
| Learn label representations from label descriptions jointly with documents and words (gamma=1/3 vs 0) (ablation) | AiTextML · BioASQ test (seen labels) | rank loss: 0.05217 -> 0.03544 | Improved |
| Learn label representations from label descriptions jointly (ablation) | AiTextML · BioASQ test (seen labels) | one-error: 0.41728 -> 0.25992 | Improved |
| Learn label representations from label descriptions jointly (ablation) | AiTextML · BioASQ test (seen labels) | average precision: 0.36645 -> 0.32786 | Hurt |
| Learn from label descriptions: effect on frequent vs rare labels (ablation) | AiTextML · BioASQ test (seen labels) | label-based average rank | Mixed |
| Learn from label descriptions (reduces overfitting) (ablation) | AiTextML · BioASQ train vs test rank loss | rank loss | Improved |
| Infer unseen label representations via joint inference step instead of averaging word embeddings of descriptions (ablation) | AiTextML · BioASQ zero-shot | rank loss: 0.52890 -> 0.21622 | Improved |
| Infer unseen label representations instead of averaging word embeddings (ablation) | AiTextML · BioASQ zero-shot | average precision: 0.00388 -> 0.02665 | Improved |
| Use label descriptions instead of label names when averaging word embeddings (ablation) | AiTextML · BioASQ zero-shot | rank loss: 0.52335 -> 0.52890 | No clear effect |
| Train word embeddings with label descriptions (AiTextML vs baseline embeddings) for averaging unseen labels (ablation) | averaged label embeddings · BioASQ zero-shot | rank loss: 0.48812 -> 0.52890 (desc.) | Hurt |
Closest papers by meaning
- Distributional Semantics Meets Multi-Label Learning (AAAI 2019 · similarity 0.87)
- Modeling Label Space Interactions in Multi-label Classification using Box Embeddings (ICLR 2022 · similarity 0.87)
- AttentionXML: Label Tree-based Attention-Aware Deep Model for High-Performance Extreme Multi-Label Text Classification (NeurIPS 2019 · similarity 0.86)
- Label-Specific Feature Augmentation for Long-Tailed Multi-Label Text Classification (AAAI 2023 · similarity 0.86)
- Compositional Generalization for Multi-Label Text Classification: A Data-Augmentation Approach (AAAI 2024 · similarity 0.85)
- Partial Multi-View Multi-Label Classification via Semantic Invariance Learning and Prototype Modeling (ICML 2024 · similarity 0.84)
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.
