Cataloguing for the open web

ALA 2014: Understanding schema.org

ALA 2014

Systems Librarian, Laurentian University

The saddest robots.txt a library catalog can offer

User-Agent: *
Disallow: /

AKA "Machines are not welcome here."

While I expect this sort of nonsense from web-hostile proprietary library systems, it was surprising to me to find that even many of the free-as-in-freedom library systems use an absolutely restrictive robots.txt file.

Libraries provide multiple access points for resources in catalogues

Users need to know that catalogues exist

Google, Bing, Yahoo are the ready-at-hand search tools

To expand access to our resources ...

We need to expose our access points to third-party search engines

Thus the sadness about absolutely restrictive robots.txt files...

and schema.org is the vocabulary initiated by the search engines

We must not let the perfect be the enemy of the good

While I appreciate the sentiment expressed by Gordon Dunsire about the importance of RDA as a vocabulary being able to provide labels for developers in any possible language without privileging any particular language, in practice this makes it incredibly difficult for most developers to actually work with.

Similarly, designing a vocabulary that can respect all cultures globally, for example audience ratings systems that classify content as suitable not just by different age ranges but also by gender, is (possibly) noble, but realistically an impossible requirement.

If you want employment for life as a vocabulary developer for a single vocabulary, RDA might be a good choice.

If you want to get things done now, and then progressively improve them, then schema.org is a good choice.

Of course they are not mutually exclusive: you will be able to publish schema.org from RDA-sourced linked data, whenever (or if) general library systems get around to offering RDA as a source.

This is what I have been doing!

Recipe for schema.org catalogue success

Allow machines to crawl your pages: robots.txt and persistent URIs
Embed machine-readable schema.org metadata into ordinary catalogue pages
Generate sitemaps for your record detail pages
Tell the search engines about the sitemaps

Unstructured -> structured -> linked data

Bibliographic records
- Record types: Book, Map, MusicRecording, fallback to CreativeWork
Holdings, using the Product/Offer model
Libraries
- Hours of operation
- Web site
- Location
- Contact information
- Branch relationships

This is the schema.org data from a library page taken from an Evergreen catalogue in the wild, as hosted by Bibliomation in Connecticut.

The library page is linked from the Holdings-as-Offer entities on the record details page; Evergreen enables you to link to your own library web site instead. But of course you can use the Evergreen-generated page as a template for ensuring that your own site contains the appropriate schema.org data.

The rendering is courtesy of RDFa Playground.

Teach your systems to speak RDFa + schema.org

Practical linked data for open source preconference
My portion was a hands-on, self-guided tutorial teaching RDFa + schema.org in library systems:
- Mark up bibliographic record detail pages
- Express library holdings as schema.org Offer entities
- Mark up library pages with opening hours, contact info, etc
- Build a simple union catalogue in five minutes using Google Custom Search Engine
- Build a proof of concept sitemap crawler / RDFa extractor

All materials available for free under a CC-BY-SA license!
Go to https://coffeecode.net
You might need to wade through a blog post or two

Addressing gaps in schema.org

In 2013 schema.org moved to W3C WebSchemas as the public forum for enhancement proposals

Enter the W3C schema.org Bibliographic Extension Community Group

AKA SchemaBibEx
A rag-tag group of linked data bibliographic enthusiasts trying to establish consensus
Using open source library systems (Evergreen, Koha, VuFind) as proofs of concept / reference implementations

Some schemabibex wins

Early: moving "citation" from MedicalScholarlyArticle to CreativeWork
Middle: Holdings-as-Offers as a best practice, with decommercialized descriptions
So close we can taste it: Periodicals and CreativeWork relationships

The future is now

Linked data is happening,
- for real,
- en masse,
- thanks in no small part to schema.org
But Jason is bringing more of the future to you...
JSON-LD and Actions and more, oh my!