Predictive and Interactive Management of Potential Inconsistencies in Business Rules

业务规则（BR）通常用于对执行的流程应遵守的公司法规或规则进行建模。实际的BR集可以手工制作，从过程观察中挖掘，或从自然语言规范中导出。由于这些都是容易出错的机制，因此所得到的BR集可能包含不一致和其他问题，这可能导致在流程执行期间根据这些规则验证流程时出现意外行为和昂贵的调试。一种具有挑战性的问题是潜在的不一致性，即，然而，只有在执行期间观察到某些事实组合时才会出现某些规则的矛盾结论。在MIB中，我们解决了BR中此类潜在不一致的预测性和交互式管理。为此，我们将利用不一致性度量（IM）和形式论证的方法，这两个方法都是知识表示（KR）和推理领域的研究课题。特别是，我们将首先建立一个正式的框架来调查潜在的不一致性的概念，调查它在最先进的规则形式主义，如DMN，Decorative和FCL，并分析计算复杂性等属性。此外，我们将开发方法来分析，可视化，特别是测量潜在的不一致性，为建模者提供对潜在问题严重性的评估。此外，使用正式论证的方法，我们将开发一个对话系统，允许建模者交互式地处理和解决规则库中的问题。这些方法将嵌入在BR基础的创作阶段，以支持BR开发周期早期的建模人员。这也包括我们的方法集成到现有的规则挖掘算法和开发新的不一致意识的。

Business rules (BR) are commonly used to model company regulations or rules that should be adhered to by executed processes. The actual set of BRs can be manually crafted, mined from process observations, or derived from natural language specifications. As these are error-prone mechanisms, the resulting set of BRs may however contain inconsistencies and other issues, which may lead to unexpected behaviour and costly debugging when processes are validated against these rules during process execution. A challenging type of issues are potential inconsistencies, i.e., contradictory conclusions of certain rules that will however only occur when certain fact combinations are observed during execution. In MIB we address the predictive and interactive management of such types of potential inconsistencies in BRs. For that, we will make use of approaches of inconsistency measurement (IM) and formal argumentation, both being research topics within the area of knowledge representation (KR) and reasoning. In particular, we will first set up a formal framework for investigating the notion of a potential inconsistency, investigate it in state-of-the-art rule formalisms such as DMN, Declare, and FCL, and analyze properties such as computational complexity. Further, we will develop approaches to analyze, visualize, and, in particular, measure potential inconsistencies to provide the modeler with an assessment on the severity of potential issues. Moreover, using methods from formal argumentation, we will develop a dialog system that allows the modeler to interactively address and solve the issues in the rule base. These methods will be embedded in the authoring phase of BR bases to support modelers early in the development cycle of BRs. This also includes the integration of our methods into existing rule mining algorithms and the development of novel inconsistency-aware ones.