Research map / Label-efficient and robust learning / Recognition and long-tail
Distinguish Polarity in Bag-of-Words Visualization
Yusheng Xie, Zhengzhang Chen, Ankit Agrawal, Alok Choudhary
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 |
|---|---|---|---|
| concatenate Jaro-Winkler literal similarity E, re-encoded by smooth AE (k=2), to word vectors | t-SNE on word vectors · Polarity vocabulary, emotion domain, M=2 | cross-entropy: .042 -> .021 | Improved |
| concatenate re-encoded literal similarity (AE k=2) to word vectors | t-SNE on word vectors · Polarity vocabulary, temperature domain, M=2 | cross-entropy: .078 -> .027 | Improved |
| concatenate re-encoded literal similarity (AE k=2) to word vectors | t-SNE on word vectors · Polarity vocabulary, size domain, M=5 | cross-entropy: .309 -> .220 | Improved |
| concatenate re-encoded literal similarity (AE k=2) to word vectors | t-SNE on word vectors · Polarity vocabulary, gender domain, M=3 | cross-entropy: .119 -> .062 | Improved |
| concatenate re-encoded literal similarity (AE k=2) to word vectors | t-SNE on word vectors · Polarity vocabulary, temperature domain, M=3 | cross-entropy: .025 -> .025 | No clear effect |
| concatenate raw [XE] without re-encoding (ablation) | t-SNE on word vectors · Polarity vocabulary, size domain, M=4 | cross-entropy: .175 -> .741 | Hurt |
| replace linear PCA by smooth AE (k=2) for re-encoding [XE] (ablation) | t-SNE on [XE] · Polarity vocabulary, emotion domain, M=4 | cross-entropy: .108 -> .035 | Improved |
| replace sparse AE (k=1) by smooth AE (k=2) (ablation) | t-SNE on [XE] · Polarity vocabulary, gender domain, M=2 | cross-entropy: .071 -> .060 | Improved |
| replace sparse AE (k=1) by smooth AE (k=2) (ablation) | t-SNE on [XE] · Polarity vocabulary, temperature domain, M=5 | cross-entropy: .060 -> .029 | Improved |
| replace linear PCA by smooth AE (k=2) (ablation) | t-SNE on [XE] · Polarity vocabulary, size domain, M=3 | cross-entropy: .178 -> .177 | Improved |
| raise concentration weight beta with k=0.5 (ablation) | AE hidden layer · enwik100 neuron activations | activation distribution | Mixed |
| raise beta with k=2 (ablation) | AE hidden layer · enwik100 neuron activations | activation distribution | Mixed |
Closest papers by meaning
- On the Dimensionality of Word Embedding (NeurIPS 2018 · similarity 0.86)
- ViCo: Word Embeddings From Visual Co-Occurrences (ICCV 2019 · similarity 0.86)
- Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models (NeurIPS 2025 · similarity 0.84)
- By Tying Embeddings You Are Assuming the Distributional Hypothesis (ICML 2024 · similarity 0.84)
- The Hidden Language of Diffusion Models (ICLR 2024 · similarity 0.84)
- On the Emergence of Linear Analogies in Word Embeddings (NeurIPS 2025 · 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.
