Research map / Label-efficient and robust learning / Recognition and long-tail
STEM: Unsupervised STructural EMbedding for Stance Detection
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 |
|---|---|---|---|
| Use STEM structural speaker embedding and 2-core classification instead of Global Embedding | CreateDebate post stance detection · CreateDebate posts | accuracy (Average): 0.75 (macro) -> 0.86 | Helped |
| Propagate labels beyond the 2-core (ablation) | STEM · CreateDebate authors | accuracy (Average): 0.88 -> 0.85 | Hurt |
| Use STEM structural speaker embedding and 2-core classification instead of Global Embedding | 4Forums post stance detection · 4Forums posts | accuracy (Average): 0.85 (macro) -> 0.89 | Helped |
| Propagate labels beyond the 2-core (ablation) | STEM · 4Forums authors | accuracy (Average): 0.79 -> 0.76 | Hurt |
| Use structural SDP embedding and hyperplane rounding rather than max-cut on the conversation tree | ConvinceMe post stance detection · ConvinceMe topics with both methods reported | accuracy: 0.84 -> 0.98 (Gay Marriage); 0.82 -> 0.99 (Evolution) | Helped |
| Classify only speakers in the 2-core rather than including propagated non-core speakers (ablation) | STEM · CreateDebate authors | accuracy (Average): 0.85 -> 0.88 | Helped |
| Classify only speakers in the 2-core rather than including propagated non-core speakers (ablation) | STEM · 4Forums posts | accuracy (Average): 0.89 -> 0.89 | No clear effect |
| Restrict classification to speakers whose vectors fall inside tighter cones | STEM on 4Forums conversations · 4Forums speakers inside the selected cone | accuracy: 0.79 -> 0.85 | Helped |
| Apply STEM to conversations with higher measured valence | STEM on 4Forums · 4Forums conversations grouped by valence | accuracy: 0.75; 0.8; 0.92 | Helped |
Closest papers by meaning
- Few-Shot Cross-Lingual Stance Detection with Sentiment-Based Pre-training (AAAI 2022 · similarity 0.86)
- The Power of LLM-Generated Synthetic Data for Stance Detection in Online Political Discussions (ICLR 2025 · similarity 0.85)
- A Topic-Aware Reinforced Model for Weakly Supervised Stance Detection (AAAI 2019 · similarity 0.84)
- MSME: A Multi-Stage Multi-Expert Framework for Zero-Shot Stance Detection (AAAI 2026 · similarity 0.84)
- Learning Shared Representations from Unpaired Data (NeurIPS 2025 · similarity 0.80)
- Multi-Modal Discussion Transformer: Integrating Text, Images and Graph Transformers to Detect Hate Speech on Social Media (AAAI 2024 · similarity 0.80)
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.