Structured data NOW

Seeding in library systems

Dan Scott / +DanScottCan
Laurentian University,

Basic premises of this talk

  1. We want our resources to be found via normal search engines
  2. We want our metadata to be readily accessible and reusable by non-library developers

Bags of words are still hard

Bags of words are still hard

Bags of words are still hard

Libraries: early technology adopters

  • Telnet access to catalogues
  • Z39.50 protocol for sharing records
  • OpenURL protocol for resolving article requests
  • COinS microformat for embedding citations in HTML
  • unAPI for offering different metadata representations

Consistent flaw: almost entirely library-specific technologies

MARC: (still) kind of a big deal

alternate title:

Tales of a semantic web dropout!

In all honesty

  • I wanted to make my library system part of the Semantic Web, but:
    • RDF/XML, N3, Turtle
    • conneg
    • Vocabularies and ontologies
    • open-world vs. closed-world assumptions
    • Triple stores, SPARQL
    • OWL, mixed specialist vocabularies
    • Billions of triples vs practical incentives

So I focused on improving Evergreen in other ways...

One day at Google I/O 2011...

Talk: Scalable structured markup (for Rich Snippets)

  • In July 2011, was announced by major search engines (Google, Yahoo, Yandex)
  • Some goals:
    1. Offer a simple vocabulary for the short tail of web results (events, products, people)
    2. Enable normals to add markup (via RDFa Lite or microdata) without having to be Semantic Web experts
    3. Enable search engines to aggregate data and apply finer-grained disambiguation and relevance strategies
  • A reaction to the state of the Semantic Web hierarchical types Creative Works Book focus on examples

Baby steps

  • By early 2012, Evergreen was publishing simplistic title / author / keyword via microdata
  • OCLC WorldCat also started publishing rich, heavily extended via JSON

Practical problems close to home

Our decentralized infrastructure has been centralized.

If your holdings are not in OCLC, you're not linked to Google Books.

One day at Google I/O 2012...

Ade Oshineye was singing the praises of Google+ Local pages

Some guy commented on Ade's G+ post about the presentation:

Still looking forward to a simple solution for automated Page updates for non-spammy metadata like hours of operation (yes, I was the guy who asked that question). Maybe we could teach G+ to scrape some selected microdata from an authoritative linked Web site, if pushing seems too dangerous?

Linked data library system leadership

  • Some library systems have supplemented or discarded MARC for a linked data model:
    • Swedish Union Catalog
    • German National Library
    • Bibliothèque nationale de France

But is your library...

  • Large (national scale)
  • Able to sustain multi-year development efforts with concurrent systems
  • Able to mandate significant changes to established practices

Better living through Web standards

Better living through Web standards

  • Hypothesis: We can iterate towards linked data with existing systems
  • Enhance our catalogues with web standards:
    • Persistent URIs
    • HTML5
    • RDFa (or microdata) expressing
    • Sitemaps listing all the URIs of interest

All of this has happened before...

  • That library standards bodies increase library participation in Semantic Web standardization, develop library data standards that are compatible with Linked Data, and disseminate best-practice design patterns tailored to library Linked Data;
  • That data and systems designers design enhanced user services based on Linked Data capabilities, create URIs for the items in library datasets, develop policies for managing RDF vocabularies and their URIs, and express library data by re-using or mapping to existing Linked Data vocabularies;

Authors included code4lib regulars such as Karen Coyle, Antoine Isaac, Peter Murray, Jodi Schneider, Ross Singer, Ed Summers, Jeff Young (apologies for anyone I've missed!)

Reality check: 2012 Common Crawl

  • Ronallo(*) found that American academic libraries published under 10,000 instances in total

RDF in Attributes (RDFa)

  • 2004 proposal that started to take off with HTML5
  • RDF serialization that decorates HTML with a dozen or so attributes
  • RDFa Lite is pared down to just five attributes
  <head><title>Fancy academic article</title></head>
  <body vocab="" typeof="ScholarlyArticle">
    <h1 property="name">The web as intellectual rising tide</h1>
    <div property="articleBody">
      <p>As my esteemed colleague
        <a href="" typeof="Person"><span
           property="name">Foo Bar</span></a> has 
        <a href="" property="citation">astutely
           observed</a> ...
  • microdata is a roughly equivalent form of inline markup

We're all in this together: SchemaBibEx

The mission of this group is to discuss and prepare proposal(s) for extending schemas for the improved representation of bibliographic information markup and sharing.
  • Founded by Richard Wallis (OCLC) in September 2012
  • Formally known as the W3C Schema Bib Extend community group
  • Participants include Karen Coyle, Ross Singer, Corey Harper, Laura Dawson, Niklas Lindstrom, Jeff Young, and many others

SchemaBibEx: Citations everywhere

  • Originally only the MedicalScholarlyArticle type had a citation property
  • Alf Eaton, via SchemaBibEx, proposed making the property applicable to all CreativeWork types
  • Status: accepted

SchemaBibEx: Library items as Offers

  • GoodRelations vocabulary for surfacing product information, based on an agent-object-promise model, was incorporated into
  • Library-specific holdings vocabulary seemed unlikely to gain traction from search engines (express library data by re-using or mapping to existing Linked Data vocabularies)
  • Libraries therefore offer items for "lease" at zero cost
  • Status: SchemaBibEx recommended practice

Mapping holdings to Offers

Offer propertyLibrary item
skuCall number
availableAtOrFromShelving location
availabilityItem status

Holdings as Offers: example

<tr property="offers" typeof="Offer">
    <a property="seller" typeof="Library"
         property="name">C.H. Booth Library - Newtown</span></a>
    <link property="businessFunction"
    <link property="itemOffered" href="#schemarecord">
  <td><span property="sku">641.5 ROM c. 2</span></td>
  <td property="serialNumber">34014012862984</td>
  <td property="availableAtOrFrom">Adult Nonfiction</td>
  <td><link property="availability"

SchemaBibEx: Periodicals

  • had an Article type but no containing publication types
  • Drove a Periodical extension proposal that supported journal/magazine relationships:
    • Article (and all child types such as ScholarlyArticle)
      • :: isPartOf :: PublicationIssue
      • :: isPartOf :: PublicationVolume
      • :: isPartOf :: Periodical
      • :: isPartOf :: Book
  • Enables structured inline citations, regardless of citation format!
  • Status: Under consideration by

SchemaBibEx: Comics

SchemaBibEx: current status

  • Exited active extension proposal development mode
  • Documenting best practices and supporting adopters
  • Will call ad-hoc meetings for new extension requirements
  • Join our community and post questions to the mailing list!

Noses to the grindstone

Tackling the problem with Evergreen

  • Simplistic microdata with the 2.2 and 2.3 releases:
    • title/author/keyword properties
    • Direct plain text values (no embedded types)
  • 2.4 release broke entities out as Person and Organization types, separated birthDate and deathDate values
  • 2.5 release (late 2013) switched to RDFa, exposed holdings as Offer types
  • 2.6 release (soon!) includes links from holdings to Library types

Further open source implementations

  • Koha 3.14 and VuFind 2.1 accepted patches to express bibliographic data + holdings
  • Independently, the Blacklight library system was enhanced to publish data
  • Approximately 4,000 library systems will publish data as sites adopt the latest releases
  • Open source library systems served as the reference implementations for our work

OCLC previews 194 million open Work descriptions

  • Huge props to OCLC Research for publishing the Open Linked Data Bibliographic Work descriptions in February 2014
  • Builds on the work of SchemaBibEx and existing proposals such as miniSKOS
  • Great steps forward!

Structured library information

Structured library information

  • offers a Library organization type
  • ILS often tracks library hours, contact information, address, branch relationships
  • Evergreen, as of 2.6 release, will expose this information by default
  • You should not have to maintain business pages in umpteen different social networks!

Structured library information example

Structured library information example

<div vocab="" typeof="Library">
    <h1 property="name">Example Branch 1</h1>
    <div><a href=""
        property="url">Library web site</a></div>

    <h2>Opening hours</h2>
        <div property="openingHoursSpecification"
            <link property="dayOfWeek"
            Monday: <span property="opens">12:00 AM</span> - 
                    <span property="closes">12:00 AM</span>
    <h2>Contact information</h2>
        <div>Email: <a href=""
        <div>Telephone: <a href="tel:(555) 555-0271"
            property="telephone">(555) 555-0271</a></div>

        <div property="location address" typeof="PostalAddress">
            <h3 property="contactType">Mailing address</h3>
            <span property="streetAddress">BR1<br>123 Main St.</span><br>
            <span property="addressLocality">Anywhere</span><br>
            <span property="addressRegion">GA</span><br>
            <span property="addressCountry">US</span><br>
            <span property="postalCode">30303</span><br>

    <h2>Branch relationship</h2>
    <div id="branch-info">Parent library: 
        <a property="branchOf"
           href="/eg/opac/library/SYS1">Example System 1</a>

Thought experiment

  • So we're exposing our holdings to the web...
  • in a common format understood by search engines...
  • linked to the library that holds the items...
  • with that library's physical address, hours of operation, web site and contact info...
  • ...
  • when can we stop doing batch uploads to OCLC and/or other union catalogues?


  • Standard documented at
  • Simple XML lists of URLs that should be crawled, with optional change information:
  • <urlset xmlns="">
  • Basic SQL for an Evergreen sitemap-generating script: Google Docs

Quick union catalogues

  • Many libraries have resource-sharing agreements, but infrastructure requires crufty Z39.50 lookups or periodic MARC batch loads
  • Promise: Common vocabulary and mapping practices should ease cross-system integration
  • Use case: Google Custom Search Engine (CSE) supports facets based on structured data such as
  • Result: It works!
    • Combined Evergreen + VuFind instances in a single search instance
    • Faceted by author value
    • Exposes Google's simplistic string-based implementation of structured data

Quick union catalogues: in progress

  • Schema UnionCat: < 100 lines of Python to crawl sitemaps and extract complex RDF
  • Currently a proof of concept
  • Future enhancements (if there is interest):
    • Only crawl new and updated sitemap entries
    • Store and search extracted RDF
    • Check item availability on demand


  • Web standards (sitemaps, RDFa, > library-niche solutions and offer a path back to decentralization
  • Catalogues can feed metadata directly to general search engines
  • Open source implementations strengthen proposals; hopefully hasten proprietary implementations
  • We have much more to do:
    • Update Common Crawl analyses: broaden domains for 2012, update for 2014
    • Enhance proof-of-concept RDFa crawler / extractor / union catalogue
    • UNIMARC and NORMARC mappings for Koha
    • Evolve MARC towards directly linked data
  • SchemaBibEx now in community support mode; if you have implementation questions, ask us!