LOCUS-Fines is a fine-grained annotation layer for the LOCUS (Local Ordinance Corpus for the United States) dataset. It annotates roughly 600,000 rows with detailed information about fines, using the gemini-3.5-flash-lite model for the annotations. The dataset was created by LocalLaws and was last updated on July 23, 2026.
Use Cases
- Extract and analyze fine amounts from legal text based on the described annotations.
- Train or evaluate models for named entity recognition of monetary penalties in ordinances.
- Study patterns and distributions of fines across different U.S. local jurisdictions.
- Perform legal compliance analysis by identifying penalty structures in municipal codes.
Strengths
- Contains annotations for approximately 600,000 rows of legal text.
- Uses an LLM (gemini-3.5-flash-lite) for annotation with a described verification step where dollar amounts are checked against original text to reduce hallucinations.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- The description notes a potential for miscategorization where numbers referring to other concepts (like months) may be incorrectly labeled as fines.
Provenance
- Source
- LocalLaws
- Collection Method
- Annotations generated using the gemini-3.5-flash-lite LLM, with verification against source text.
- Time Range
- null
- Freshness
- Last updated 2026-07-23 04:43:18; freshness should be verified.
- Geography
- United States (local ordinances)