Developing Cyberinfrastructure Tools and Services for Metadata Quality Evaluation

Metadata and data quality are at the core of reusable and reproducible science. While great progress has been made over the years, much of the metadata collected only addresses data discovery, covering concepts such as titles and keywords. Improving metadata beyond the discoverability plateau means documenting detailed concepts within the data such as sampling protocols, instrumentation used, and variables measured. Given that metadata commonly do not describe their data at this level, how might we improve the state of things? Giving scientists and data managers easy to use tools to evaluate metadata quality that utilize community-driven recommendations is the key to producing high-quality metadata. To achieve this goal, we created a set of cyberinfrastructure tools and services that integrate with existing metadata and data curation workflows which can be used to improve metadata and data quality across the sciences. These tools work across metadata dialects (e.g., ISO19115, FGDC, EML, etc.) and can be used to assess aspects of quality beyond what is internal to the metadata such as the congruence between the metadata and the data it describes. The system makes use of a user-friendly mechanism for expressing a suite of checks as code in popular data science programming languages such as Python and R. This reduces the burden on scientists and data managers to learn yet another language. We demonstrated these services and tools in three ways. First, we evaluated a large corpus of datasets in the DataONE federation of data repositories against a metadata recommendation modeled after existing recommendations such as the LTER best practices and the Attribute Convention for Dataset Discovery (ACDD). Second, we showed how this service can be used to display metadata and data quality information to data producers during the data submission and metadata creation process, and to data consumers through data catalog search and access tools. Third, we showed how the centrally deployed DataONE quality service can achieve major efficiency gains by allowing member repositories to customize and use recommendations that fit their specific needs without having to create de novo infrastructure at their site.