Scary Metadata!

One of the best times to do metadata remediation is during a platform migration, but it can be a scary undertaking. Don’t fret! Searching the dark corners of your digital repository allows you to shed light on what you need to fix and what can be improved.

One anecdote from Anne Washington, Metadata Services Coordinator at University of Houston, is doubly-spooky! Graphic content and editorializing in the metadata, the horror!

I’ll share an example. We have a digitized film from the 1960s on chemical and biological warfare in which scientists demonstrate the effects of chemical and biological weapons on a rabbit. Disturbing. The description originally created contained some editorializing: “A perfectly healthy rabbit was exposed to a chemical by the scientist and within a few minutes it no longer seem to be moving.(Poor Rabbit!!!)” and the subjects assigned focused primarily on animal welfare. And while this is certainly a component of this film, it was clear that the metadata creator had a particular view on this subject.

While neutrality in description is impossible, as metadata specialists, our current general practice is to try to describe resources as best we can without adding interpretation (unless it is provided to us by a collaborator). In order to catch these types of issues in legacy collections, we have to re-analyze the item, its description, and subject headings; no small task. So, often this is not something we can focus on during large scale metadata cleanup. We do the best we can, but sometimes we miss things and luckily our users share feedback that draws our attention to these kind of issues.

When Dean Seeman, Head of Metadata at University of Victoria Library, did his first round of metadata remediation he found many a lone skeleton:

[out of] 162 different metadata fields/properties 89 were used only once!

I think this is a result of at least a couple of things: lack of centralized direction over metadata in the system as a whole (and lack of communication around consistent standards) ; and CONTENTdm allowing anyone to name their properties/field labels anything they want.

As a result we had 4 different properties/field labels in CONTENTdm of expressing Alternative Title:
Title - Alternative (used in 14 collections)
Alternative Title (11 collections)
Title-Alternative (4 collections)
Title, Alternate (2 collections)

Another example of this is around Rights field labels:
Access and Use Rights (27 collections)
Use Rights (19 collections)
Access Rights (17 collections)
Rights (5 collections)
Use and Access Rights (27 collections)

Our path to Hyku involves creating a single application profile (learn more here ) for all collections which will force us to standardize our language around property labels and also take some of the pressure/ambiguity out of the decision of “what fields to use.” Just one example of the opportunities that a migration affords.

As Dean and Anne state above, learning about your scariest metadata issues is a great way to start the journey towards cleaner and more standardized metadata.