Research map / Label-efficient and robust learning / Recognition and long-tail
Active Discriminative Text Representation Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| select sentences by maximum expected gradient on constituent word embeddings | CNN sentence classification · MR, CR, and Subj | AUC: 0.690 -> 0.707; 0.732 -> 0.743; 0.840 -> 0.856 | Helped |
| use expected gradients on softmax parameters instead of word embeddings (ablation) | CNN sentence classification · MR, CR, and Subj | AUC: 0.707 -> 0.667; 0.743 -> 0.674; 0.856 -> 0.785 | Hurt |
| replace the maximum word-gradient score with an average over word gradients (ablation) | EGL-word sentence selection · Sentence classification | AUC | Hurt |
| combine document word-gradient scores with entropy using a time-varying stochastic interpolation (ablation) | CNN document classification · MR, DR, and MuR | AUC: 0.719 -> 0.725; 0.889 -> 0.893; 0.718 -> 0.736 | Helped |
| use word-gradient scoring without entropy interpolation (ablation) | CNN document classification · MR document classification | AUC: 0.719 vs. 0.719 | No clear effect |
| focus only on word-gradient scoring rather than shifting toward instance uncertainty (ablation) | EGL-word-doc document selection · Document classification over learning iterations | accuracy | Mixed |
| draw the interpolation weight from a distribution rather than set it deterministically (ablation) | EGL-Entropy-Beta · Document classification | accuracy | Helped |
Closest papers by meaning
- A Simple but Tough-to-Beat Baseline for Sentence Embeddings (ICLR 2017 · similarity 0.90)
- Improving Word Embeddings with Convolutional Feature Learning and Subword Information (AAAI 2017 · similarity 0.90)
- A New Method of Region Embedding for Text Classification (ICLR 2018 · similarity 0.89)
- Repurposing Language Models into Embedding Models: Finding the Compute-Optimal Recipe (NeurIPS 2024 · similarity 0.88)
- Contextual Document Embeddings (ICLR 2025 · similarity 0.87)
- Attention with Trained Embeddings Provably Selects Important Tokens (NeurIPS 2025 · similarity 0.86)
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.