Al Rcta: Real-World Crowdsourced Text Annotations for Active Learning Analysis

Three benchmark text-classification datasets support a study analyzing active learning algorithms. The data includes cleaned datasets, collected annotator files, and experimental splits from a recently accepted paper. Author varuntotakura published this repository on Hugging Face in May 2026.

Use Cases

• Benchmarking active learning algorithms based on the three provided text-classification datasets
• Reproducing experimental splits for active learning studies based on the included scripts
• Analyzing annotator behavior and crowdsourcing patterns based on the collected annotator files
• Training text classifiers for news categorization based on the AG News subset

Strengths

• Accompanies a peer-reviewed, recently accepted paper, suggesting academic rigor.
• Includes scripts to reproduce the experimental splits, supporting reproducibility.
• Contains three established benchmark datasets, including AG News for text classification.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale experiments.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Paper and repository by author varuntotakura.

Collection Method

Crowdsourced text annotations, cleaned and prepared for research.

Freshness

Last updated 2026-05-12 02:48:20; freshness should be verified.