Research map / Image generation / Super-resolution
Predicting Demographics of High-Resolution Geographies with Geotagged Tweets
Omar Montasser, Daniel Kifer
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 |
|---|---|---|---|
| Use the proposed Normalized User + Gaussian configuration instead of the adapted prior-work baseline | Demographic count prediction with unknown population size · Block, blockgroup, tract, and county | Pearson r: 2.84 (block), 1.41 (blockgroup), 1.31 (tract), and 1.21 (county) | Improved |
| Apply feature transformations, including Gaussian, Logistic, and Anscombe, to normalized bag-of-words features | Proposed method with unknown population size · Across resolutions and gender/race tasks | Pearson r: 1.88 | Improved |
| Use Raw User + TFIDF rather than national-level demographic proportions | Proposed method with known population size · Race/ethnicity, block, blockgroup, tract, and county | Pearson r: 1.15 (block), 2.05 (blockgroup), 2.42 (tract), and 1.12 (county) | Improved |
| Use the proposed model rather than the adapted prior-work baseline | Race/ethnicity prediction with known population size · Block, blockgroup, tract, and county | Pearson r: 1.01 (block), 1.07 (block group), 1.11 (tract), and 1.05 (county) | Improved |
| Use a learned demographic model instead of national proportions | Known-population gender count prediction · Block, blockgroup, tract, and county | Pearson r | No clear effect |
| Use a feature transformation on normalized bag-of-words instead of plain bag-of-words (ablation) | Proposed method with unknown population size · Across resolutions | Pearson r | Improved |
| Use User Normalization rather than Word Normalization (ablation) | Proposed method with unknown population size · Across resolutions | Pearson r | Improved |
| Use Raw User + TFIDF instead of User Normalization for known-population race prediction (ablation) | Proposed method · Race/ethnicity, block, blockgroup, tract, and county | Pearson r | Improved |
| Train for more epochs, for example 100 rather than 10 (ablation) | SGD linear regression model · Demographic prediction experiments | R² | No clear effect |
| Change the learning-rate decay parameter ρ (ablation) | SGD linear regression model · Demographic prediction experiments | R² | No clear effect |
Closest papers by meaning
- GeoLLM: Extracting Geospatial Knowledge from Large Language Models (ICLR 2024 · similarity 0.83)
- Large Language Models are Geographically Biased (ICML 2024 · similarity 0.80)
- Valid Inference with Imperfect Synthetic Data (NeurIPS 2025 · similarity 0.80)
- SPATIALEPIBENCH: Benchmarking Spatial Information and Epidemic Priors in Forecasting (NeurIPS 2026 · similarity 0.80)
- MotifGPL: Motif-Enhanced Graph Prototype Learning for Deciphering Urban Social Segregation (AAAI 2025 · similarity 0.79)
- Finetuning Text-to-Image Diffusion Models for Fairness (ICLR 2024 · similarity 0.79)
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.
