Requirements and Services

Knowledge-intensive applications pose new challenges to metadata management, including distribution, access control, uniformity of access, and evolution in time. The authors identify general requirements for metadata management and describe a simple model and service that focuses on Resource Description Framework (RDF) metadata to address these requirements. I n important application domains, dataand service providers are increasing-ly making their resources publicly available to the community for re-use in complex workflows. To make those resources useful in practice, however, providers must also provide annotations that describe the data and services ’ nature and function. This is the case for e-sci-ence, the realm of in silico experiments or “procedures that use computer-based information repositories and computa-tional analysis tools to test a hypothesis, derive a summary, search for patterns, or demonstrate a known fact.”1 In e-science, it’s quite common for providers and consumers to independent-ly add annotations to resources to facili-tate their discovery or to record details of their use as part of an experiment. Thus, when scaled to hundreds or thousands of resources and users of those resources, the annotations themselves will form a new and large corpus of heterogeneous meta-data distributed over many organizations, with no central control over its mainte-nance. As a new and complex type of data resource, such metadata requires some form of management to be of any practical use. In this article, we present a middle-ware service for metadata management that addresses the issue. Its design is based on the observation that, regardless of their differences in format and content, two simple properties are common to all metadata: namely, that it’s invariably associated to some underlying resource, and, optionally, separate meta-informa-tion for interpreting metadata — an ontol-ogy, for instance — might be available. We refer to such meta-information as a knowledge entity to underline the fact that it’s used to interpret the metadata. This i