SIRNA: Using SMT Solvers to Verify Business Rule Consistency
SIRNA, a novel tool, seeks to align the natural language descriptions of business rules with their actual implementations. Researchers introduced this system on arXiv (2608.00396), which merges large language models (LLMs) with formal verification techniques, particularly SMT solvers. SIRNA translates business rule documentation into potential SMT formulas using LLMs, subsequently validating these translations. It also transforms corresponding business rules into SMT equivalents and verifies them against the natural language representations. A case study focusing on cost calculations within tax domains illustrates a notable decrease in false positives when compared to baseline assessments. This method can be applied across various fields where business logic is documented and coded.
Key facts
- SIRNA is a tool for checking consistency between natural language business rules and their implementations.
- It uses SMT solvers and large language models (LLMs).
- The system translates natural language documentation into candidate SMT formulas.
- It validates translations and checks corresponding business rules against them.
- The case study focuses on cost calculations in tax domains.
- SIRNA reduces false positives compared to baseline evaluations.
- The method is generalizable to other domains.
- The paper is available on arXiv with identifier 2608.00396.
Entities
Institutions
- arXiv