Combining acquisition and debugging of business rule models
Résumé
Business rules (BR) can be acquired from complex texts such as laws, regulations or contracts. However knowledge extraction and formalization is a complex task that involves business experts as well as Information Technology engineers and that is error-prone. Instead of waiting until the rule base is completed or the BR decision system is put into production to detect problems, we propose to detect inconsistencies and errors at an early stage, before the formalization work is completed. This paper presents the quality procedures that can be implemented in the process of BR acquisition from NL regulations. We show that the documented business rule models under construction are useful to detect potential anomalies at a semi-formal level of the BR base, where the rules exploit a formal vocabulary but are simply structured into premises and conclusions. Even at the prior and textual level, these documented models give the business experts a global and structured view over the NL regulation, which helps the formalization process.