Research map / Label-efficient and robust learning / Recognition and long-tail
Collective Supervision of Topic Models for Predicting Surveys with Social Media
Adrian Benton, Michael Paul, Braden Hancock, Mark Dredze
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 |
|---|---|---|---|
| add collective supervision (survey values) to topic model instead of unsupervised LDA | Upstream-ada-words · Guns/Vaccines/Smoking | RMSE: 17.44 -> 11.47 (Guns) | Improved |
| use upstream (DMR) instead of downstream (SLDA) supervision (ablation) | topic models · Guns/Vaccines/Smoking | RMSE: 5.82 -> 11.22 (Vaccines, Upstream-ada vs SLDA, error lower upstream) | Improved |
| collective response variable instead of per-document response in SLDA (ablation) | Downstream SLDA · Guns/Vaccines/Smoking | RMSE: 11.22 -> 9.17 (Vaccines); 3.95 -> 4.35 (Smoking) | Mixed |
| add adaptive supervision to upstream model (ablation) | Upstream · Guns/Vaccines/Smoking | RMSE/perplexity: 15.37 -> 11.48 (Guns) | No clear effect |
| add structured word priors to upstream model (ablation) | Upstream · Guns/Vaccines/Smoking | RMSE/perplexity: 15.37 -> 11.50 (Guns) | No clear effect |
| use Census demographic supervision instead of survey supervision (ablation) | upstream/downstream models · Guns/Vaccines/Smoking | RMSE | Hurt |
| add topic model with collective supervision vs bag of words | Upstream-ada-words · Universal background checks | RMSE (2001 Y omitted): 7.31 -> 5.48 | Improved |
| train topic model on related 2001 firearm data instead of target UBC data (ablation) | Upstream-ada-words · Universal background checks | RMSE: 5.11 -> 6.66 | Hurt |
Closest papers by meaning
- Learning Supervised Topic Models for Classification and Regression from Crowds (TPAMI 2017 · similarity 0.85)
- MindVote: When AI Meets the Wild West of Social Media Opinion (AAAI 2026 · similarity 0.84)
- In Search of Coherence and Consensus: Measuring the Interpretability of Statistical Topics (JMLR 2018 · similarity 0.83)
- Learning to Curate Context: Jointly Optimizing Retrieval and Prediction for Multimodal Social Media Popularity (AAAI 2026 · similarity 0.83)
- Improving Multimodal Social Media Popularity Prediction via Selective Retrieval Knowledge Augmentation (AAAI 2025 · similarity 0.82)
- SMTPD: A New Benchmark for Temporal Prediction of Social Media Popularity (CVPR 2025 · similarity 0.81)
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.
