Evaluating and Evolving Metadata in Multiple Dialects

Despite many long-term homogenization efforts, communities continue to develop focused metadata standards along with related recommendations and (typically) XML representations (aka dialects) for sharing metadata content. Different representations easily become obstacles to sharing information because each representation generally requires a set of tools and skills that are designed, built, and maintained specifically for that representation. In contrast, community recommendations are generally described, at least initially, at a more conceptual level and are more easily shared. For example, most communities agree that dataset titles should be included in metadata records although they write the titles in different ways. This situation has led to the development of metadata repositories that can ingest and output metadata in multiple dialects. As an operational example, the NASA Common Metadata Repository (CMR) includes three different metadata dialects (DIF, ECHO, and ISO 19115-2). These systems raise a new question for metadata providers: if I have a choice of metadata dialects, which should I use and how do I make that decision? We have developed a collection of metadata evaluation tools that can be used to evaluate metadata records in many dialects for completeness with respect to recommendations from many organizations and communities. We have applied these tools to over 8000 collection and granule metadata records in four different dialects. This large collection of identical content in multiple dialects enables us to address questions about metadata and dialect evolution and to answer those questions quantitatively. We will describe those tools and results from evaluating the NASA CMR metadata collection.