基于CWM的元数据的形式化推理框架研究

A Formal Framework for Reasoning on Metadata Based on CWM

在基于公共仓库元模型(CWM)建立元数据的过程中,参与建立元数据的团体的不同经验以及描述数据的不同视角不可避免地带来元数据的冲突和冗余等问题,然而CWM的图形化特点使之缺乏精确的语义,所以如何在其上进行推理以自动发现这些问题至今没有得到很好的解决.研究了利用描述逻辑——一个一阶谓词逻辑的可判定子集形式化CWM元模型和元数据并进行推理的方法,将一致性检测分为水平一致性和演化一致性分别处理,在处理演化一致性的过程中对CWM元模型进行了扩展,使之支持元数据的版本能力从而能够推理由于演化引起的不一致问题,然后利用推理引擎LOOM对两种情形进行推理检测以发现不一致信息,取得了令人满意的结果,表明提出的方法是可行的.

During the metadata creation based on common warehouse metamodel （CWM）, the different experiences and views of describing data of organizations involved in metadata creation cause some problems inevitably, such as inconsistencies and redundancies. However, reasoning on CWM metadata for automatically detecting these problems is difficult because CWM metamodel and metadata are rendered to users by graphs, which lack precise semantics. In this paper, an approach is proposed to formalize and reason on CWM metamodel and metadata in terms of a logic belonging to description logics, which are subsets of first-order logic. First, according to the specialities of CWM metamodel and metadata, description logic DLid which supports identification constraints on concepts is proposed. Then consistency is distinguished into horizontal consistency and evolution consistency. Towards evolution consistency, CWM metamodel is extended with version capabilities so that reasoning about inconsistency caused by evolution can be done and then the formalization of CWM metamodel and metadata into DLid knowledge base for horizontal consistency checking and evolution consistency checking is studied. Finally, reasoning engine LOOM is applied to check consistency for the above two situations, and the results are encouraging. The approach can be exploited for developing intelligent system that supports automated reasoning on CWM metadata, so as to provide support for the development of the components of data warehouse systems, thus improving the reliability of metadata integration and data warehouse system.