Abstract Multi-context systems introduced by Brewka and Eiter provide a promising framework for interlinking heterogeneous and autonomous knowledge sources. The notion of diagnosis has been proposed for analyzing inconsistency in multi-context systems, which captures a pair of subsets of bridge rules of a multi-context system needed to be deactivated and activated unconditionally, respectively, in order to restore the consistency for that system. Generally, diagnoses need to be prioritized from some specific perspectives in order to select the most desirable ones to resolve inconsistency. In this paper, we propose a series of intrinsic approaches to prioritizing diagnoses based on the structure of information exchange over contexts in a multi-context system, which allow us to rank diagnoses in cases where no external knowledge is available. We use a heterogeneous graph, termed information exchange network, to formulate the information exchange over contexts via bridge rules in a multi-context system. Then we propose three kinds of approaches to prioritizing diagnoses based on the information exchange network for a multi-context system. Approaches of the first kind focus on ranking diagnoses by evaluating the impact of potential revision according to each diagnosis on direct information exchange over contexts from different perspectives, whilst approaches of the second kind rank diagnoses by evaluating the impact of potential revision according to each diagnosis on betweenness centralities of contexts with regard to the whole information exchange network. The last one uses the betweenness centralities of bridge rules in the information exchange network to prioritize diagnoses directly.

中文翻译：

---

在多上下文系统中优先考虑诊断的内在方法

摘要 Brewka 和 Eiter 引入的多上下文系统为互连异构和自主知识源提供了一个有前途的框架。诊断的概念已被提出用于分析多上下文系统中的不一致性，它捕获需要分别无条件停用和激活的多上下文系统的一对桥接规则子集，以恢复该系统的一致性. 通常，需要从某些特定的角度对诊断进行优先级排序，以便选择最理想的诊断来解决不一致问题。在本文中，我们提出了一系列基于多上下文系统中上下文信息交换结构来确定诊断优先级的内在方法，这使我们能够在没有外部知识的情况下对诊断进行排序。我们使用异构图，称为信息交换网络，通过多上下文系统中的桥接规则来制定上下文信息交换。然后，我们针对多上下文系统提出了三种基于信息交换网络的优先诊断方法。第一类方法侧重于对诊断进行排序，根据每个诊断评估潜在修订对从不同角度的上下文直接信息交换的影响，而第二类方法通过根据每个诊断评估潜在修订的影响来对诊断进行排序。关于整个信息交换网络的上下文的中介中心性。