Skip to:

Making Metadata work (ISKO June 2014)

Subject: 

David Penfold pointed out that the title of this one-day ISKO course (June 23, 2014) was ambiguous: was “making metadata work” about giving ourselves a lot of work to do as we try to sort out metadata? Or was it bringing metadata under control so it actually did something useful? Either way, there was the worrying potential for metadata to be something of a pain.

If this one-day event represented the state of the art for metadata, then it was a pretty subdued state! Given the stellar line-up of speakers, I expected rather more. We were promised a workshop in the morning, and a full conference in the afternoon. What we got was often not very revealing to the uninitiated (given that many speakers had just 15 minutes), and at times made you realise more about how metadata doesn’t work than about how it does. Or about what hasn’t been cracked, such as the speaker who excited us all by the revelation that episodes of BBC’s The Archers had been indexed for the presence or absence of crying. Unfortunately, when someone asked how it had been done, he revealed that it hadn’t been done. It was just an interesting idea.

Before we get to the drawbacks, were there any highlights?

Linked open vocabularies: this looks to be an excellent initiative for using multiple public vocabularies for our metadata.
Europeana revealed they only accepted metadata that could be shared. Such a simple principle revealed something quite powerful: the onus on checking the rights for metadata is thereby passed to the institution delivering the data, which is as it should be.
Stella Dextre Clark pointed out that thesauri used to be standalone things, but were now expected to be interoperable. This has always been true of websites, but for thesauri it still has an air of something radically innovative.
Using schema.org enables us to send emails with formatting based on the structured data within them.
When data is linked, you may well not have any idea what use has been made of your metadata, which makes it difficult to know how to structure your metadata in the future to meet user needs.
Hospitals live with a kind of dual world: some call it cancer, others call it oncology. This shouldn’t be a problem, except that a search on one term in some health sites produces very different results to a search on the other.

But there were also some less successful aspects to the day:

Some of the presentations ignored what was known to others. The keynote presentation, on the importance of metadata for high quality search, was preaching to the converted. Nobody in the room would have questioned the importance of metadata, and although the presentation by Martin White was of his usual high quality, the time could have been better spent considering how to implement some of the improvements to search that are so clearly needed.
In the morning, the presentations were too short, and so became at times unintelligible.
We were sometimes told about the problems before (or without) being told the solution that was being attempted.
It would have been interesting to compare differences between (say) schema.org and RDFa, or to explore exactly what SKOS-XL is.
General advice such as, when building a thesaurus, “strike the right balance between your own market and the outside world” is fine in principle but meaningless without some examples.
Other presentations launched straight into controversies that I suspect most of us in the audience would be hard-pressed to comment on. If I am told that ISO15964 does not have a top-level array organisation, I am prepared to believe it, but unless someone explains to me why it is important, I will be none the wiser.
When linking taxonomies, often getting the match 80% right is sufficient. This sounded fascinating, but I would want to know more about how and why that percentage (rather than 100%) was enough.

My recommendation: more detailed case studies, including proper introductions and settings, would have been of more value for this subject.