Towards a framework for writing executable natural language rules

The creation of domain-specific data validation rules is commonly performed by the relevant domain experts. Such experts are often not acquainted with the low-level technologies used to actually execute these rules and will hence document them in some informal form, such as in natural language. In order to execute these rules, they need to be transformed by technical experts into a relevant executable language, such as SQL. The technical experts in turn are often not familiar with the business logic these rules are depicting and will thusly have to collaborate with the business experts to gain insight into the semantics of the rules. This paper presents an approach for writing financial data validation rules in constrained natural language, that can then be automatically transformed and executed against the data they are referring to. In order to achieve this, we use the Xtext framework for creating the editor where business experts can create their rules that can then be transformed into executable constraints. We evaluate this approach in terms of its extensibility, coverage and verboseness with respect to the business rules sent to specific UK banks submitting data under one of the Bank of England’s annual reviews.