arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1107.2088cs.AI
收藏

通过不一致性管理推进多上下文系统

Advancing Multi-Context Systems by Inconsistency Management

Vienna University of Technology(维也纳工业大学)

Antonius Weinzierl

摘要:

多上下文系统是一种表达形式主义模型(可能)异构知识库之间的非单调信息交换。然而,这样的信息交换往往会带来不可预见的副作用,导致违反约束,使系统不一致,从而无法使用。虽然有许多方法来评估和修复一个单一的不一致的知识库,多上下文系统的异质性提出了尚未以令人满意的方式解决的问题:如何识别和解释一个不一致,蔓延在多个知识库与不同的逻辑形式(例如,逻辑程序和本体论)?如果推理/信息交换是非单调的(例如,信息缺失是原因)。如果对知识库的访问受到限制,如何处理不一致性(例如,各公司交换信息,但不允许对其知识库进行任意修改)?许多传统的方法只针对一致的系统,但不一致的自动删除并不总是可取的。因此,必须支持人类操作员找到导致不一致的错误部件。在我的论文中,这些问题将主要从基础的角度来解决,而我们的研究项目也提供了算法和原型实现。

英文摘要:

Multi-Context Systems are an expressive formalism to model (possibly) non-monotonic information exchange between heterogeneous knowledge bases. Such information exchange, however, often comes with unforseen side-effects leading to violation of constraints, making the system inconsistent, and thus unusable. Although there are many approaches to assess and repair a single inconsistent knowledge base, the heterogeneous nature of Multi-Context Systems poses problems which have not yet been addressed in a satisfying way: How to identify and explain a inconsistency that spreads over multiple knowledge bases with different logical formalisms (e.g., logic programs and ontologies)? What are the causes of inconsistency if inference/information exchange is non-monotonic (e.g., absent information as cause)? How to deal with inconsistency if access to knowledge bases is restricted (e.g., companies exchange information, but do not allow arbitrary modifications to their knowledge bases)? Many traditional approaches solely aim for a consistent system, but automatic removal of inconsistency is not always desireable. Therefore a human operator has to be supported in finding the erroneous parts contributing to the inconsistency. In my thesis those issues will be adressed mainly from a foundational perspective, while our research project also provides algorithms and prototype implementations.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑