Scientific Metadata: From Filesystem to Metadata Catalog

A common problem across scientific domains concerns metadata: Important information about experiments can only be found in file names and directory structures scattered around the filesystem. Fully manual or semi-automated naming of the files, or retrospective changes in the templated structure can worsen the situation by introducing inconsistencies and ambiguities. Not only can this impede scientists in their daily work, but it also prevents reuse or automated post-processing of the data. This talk will present an approach to (i) extracting the metadata from a large collection of real-world datasets suffering from these issues and (ii) improving their (machine-)actionability by ingesting the results into a dedicated metadata catalog.