USClassActions: 3,000 U.S. Federal Court Complaints with Win/Lose Judgments

USClassActions is an English dataset of 3,000 complaints from the U.S. Federal Court with binarized judgment outcomes (Win/Lose). The dataset, created by darrow-ai and last updated on 2024-01-24, is intended to promote robustness and fairness studies in legal NLP.

Use Cases

• Train text classification models based on legal complaint documents.
• Study fairness and bias in algorithmic legal outcome prediction based on complaint text.
• Benchmark NLP model robustness on a specialized legal domain task.

Strengths

• Contains 3,000 text documents.
• Includes a binarized target variable (Win/Lose) for supervised learning.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but dataset size and file formats are unspecified.
• Freshness should be verified as the last update was 2024-01-24.

Provenance

Source

darrow-ai via Hugging Face

Collection Method

Likely collected from U.S. Federal Court records.

Time Range

null

Freshness

Last updated 2024-01-24 10:00:39

Geography

United States