Library Linked Data

LLD Incubator Group

LLD Incubator Group

• Chartered for one year from May 2010
  • Extended until August 2011
• Circulation of final report for public comments imminent
• Participation from W3C, VU Amsterdam, INRIA, Library of Congress, JISC, Deutsche Nationalbibliotek, DERI Galway, OCLC, Talis, LANL, University of Applied Sciences Potsdam, Helsinki University of Technology, University of Edinburgh and others

Preview of Final Report

• What is a Library?
• What is Library Data?
• Benefits of Linked Data
• Relevant technologies
• Implementation challenges and barriers
• Recommendations

What is a library?

• A library is...
  • A collection of materials with information content
  • A place where the collection is located
  • An agent or organisation which curates the materials and maintains the location
  • A closely related concept to "archive" and "museum"
    • Collectively, "memory institutions"
  • Dedicated to preserving and managing access to the materials over very long timescales (centuries or even millenia).

What is library data?

• Datasets: structured descriptions of "things" such as books, articles, recordings, often in the form of a catalog or set of metadata records
• Element sets: standard sets of properties or attributes and classes or entity types used to describe items in datasets
• Value vocabularies: sets of concepts such as topics, art styles, authors, publishers. Thesauri, code lists, classifications schemes, subject headings controlled vocabularies, authority records

Benefits of Linked Data

Example:

Maxwell, James Clerk, 1831-1879

• How can one refer to this person? A descriptive string? Or...

  http://dbpedia.org/resource/James_Clerk_Maxwell

  • name: James Clerk Maxwell
  • birthDate: 1831-06-13
  • birthPlace: Edinburgh, Scotland
  • deathDate: 1879-11-05
  • deathPlace: Cambridge, England
  • depiction: http://.../maxwell.png
  • knownFor: Maxwell's Equations, Maxwell's Demon, etc.

http://dbpedia.org/resource/A_Treatise_on_Electricity_and_Magnetism

dc:creator

http://dbpedia.org/resource/James_Clerk_Maxwell

Benefits of Linked Data

• Shareable

  • Globally unique resolvable identifiers - URI
  • Libraries can make trusted metadata descriptions for common use

• Extensible

  • "Open world" - no description is complete, anybody can add

    descriptive information from within their own publishing space

• Re-usable

  • Descriptions from diverse sources talking about the same thing
  • Annotations, enrichments, etc.

• Internationalisable

  • Full support for translations of terms to other languages
  • Natural language strings are not used as identifiers

Benefits of Linked Data

... to cultural institutions

• use of mainstream technologies rather than specialised formats and Integrated Library Systems specific to libraries
• sharing data, particularly using common trusted identifiers for items/works and authority data for e.g. authors, means less duplication of effort, lower infrastructure costs
• clarification of licensing of descriptive metadata
• greater visibility on the web and reuse
  • "The best thing to do with your data will be thought of by somebody else"

Benefits of Linked Data

... to librarians, archivists and curators

• use of web-based identifiers makes resources immediately available and up-to-date

• pull together descriptions for resources outside their domain environment

  • across cultural heritage datasets
  • from the web at large

• concentrate on their domain of local expertise rather than re-creating existing descriptions already elaborated by others

Benefits of Linked Data

... to developers

• use of well known standard protocols and techniques instead of domain-specific software
  • HTTP instead of Z39.50
  • RDF instead of MARC or EAD
  • REST
• freely mix or mash-up data from libraries with other sources

Available Vocabularies and Datasets

Element Sets

• Dublin Core
  • standard elements such as creator and date
• FRBR
  • entities like Work and Manifestation and elements that link and describe them
• BIBO
  • entities like Book, Article, Collection and elements like isbn, issn
• MARC21
  • a wide range of elements (fields)
• FOAF and ORG
  • elements to describe people and organisations

Value Vocabularies

• Library of Congress Subject Headings - topics of books
• Art and Architecture Thesaurus - art styles
• Virtual International Authority File - consolidated authority information
• Gemeinsame Normdatei (GND) - german names, subject headings and uniform titles
• GeoNames - geographical information

Datasets

• Project Gutenberg - 36,000 catalogue records + full text
• Talis Open Library Contribution - 5 million MARC records from various sources
• Open Library - 30 million catalogue records and over 1 million ebooks
• Crossref / DOI - 40 million academic articles and many journals
• British National Bibliography - 3 million catalogue records in RDF
• HBZ Linked Open Data - 5 million records from the Nordrhein-Westfahlen union catalogue
• Medline - 20 million scientific articles
• SUNCAT - union catalogue of uk academic journal holdings (coming soon)

Relevant Technologies

• Discrete and bulk access to information
  • discrete means individual resources with Cool URIs
  • bulk means entire datasets available for download
• Linked data front-ends to existing stores
  • D2R style mappings to relational database schemata
  • rewrite rules for SRU databases

Tools for data designers

• OWL can provide a way to do ontology mapping for working with data expressed using different element sets

  • tools like Protege can help here

• SKOS is designed for knowledge organisation, expressing taxonomies and concept schemes and the like - this is a natural fit for many value-vocabularies

  • tools like iQvoc Protege, Pool Party

Web Application Frameworks

• Systems for building specialised dynamic web sites (Django, Ruby on Rails, etc).
• Model-View-Controller arrangement (in theory)
• Usually includes URL routing
  • encourages Cool URIs
  • provides much of the scaffolding for linked-data - HTML for humans, RDF for machines from the same URI
• Can simply generate RDF data on the fly in the same way as is done with HTML, XML, JSON, etc.
  • don't throw away existing work

Microformats and Microdata

• Embed structured data in HTML

• Microdata works with HTML5, largely supplants microformats

  • http://schema.org/ joint project by Google, Microsoft Yahoo
    • Primarily for Search Engine Optimisation
    • in principle extensible but not well set up for library data as it stands
    • interoperable at a certain level with RDF (http://schema.rdfs.org/)

RDFa

• RDFa embeds full RDF into XHTML
• Data is interleaved with presentation markup
• full expressive power of RDF
• can coexist with microdata (modulo strictness about XHTML vs. HTML5)
• can be output by content management systems, e.g. Drupal

Web Services for LLD

• In theory pure linked-data approach is possible
  • linked data URIs
  • "native" rdf storage in a triple/quad store
  • SPARQL query and update language
  • inferencing with OWL description logics or other rules
• In practice...
  • this means minimal re-use of existing infrastructure
  • processing and i/o intensive compared to e.g. relational databases
  • steep learning curve for many web developers

Web Services for LLD

• OpenSearch - a convention for search interfaces
• Linked Data API - arranges access to linked data resources and sparql endpoints in a way more familiar to web developers
• JSON encodings for linked data
• ATOM and RSS also commonly used
• Ongoing work of W3C RDF Web Applications Working Group

Implementation Challenges

• Designed for stability, the library ecosystem resists change
  • stability and reliability are long term goals; libraries and
  • cultural heritage institutions are predisposed to conservatism
  • when most information was in print format, libraries were at the forefront of using computers for organisation and retrieval
  • information natively in digital form eventually overtook print, libraries have struggled to keep up

Metadata sharing

• Sharing of library data traditionally happens amongst libraries, not with the wider world
• Coorperative metadata creation is economical but centralised
  • changes to methodologies require coordination and strong leadership
  • the broader the scope of change (e.g wholesale adoption of FRBR or moving to library linked data) the more coordination and leadership is required, taking into account local needs

Staffing and Expertise

• Libraries are generally understaffed in the technology area
  • few software developers overall
  • library workers' training does not include a strong component of software development including metadata
• Technical expertise lies mostly with a small number of commercial software vendors
  • or in large academic libraries

Rapid Technological Change

• Technological environment continually evolved since the 1960s
• Tendency to engage only with well-established standards and practices
  • linked data is relatively new
  • experimentation with linked data limited partially due to staffing/expertise and partially due to lack of specific library data oriented tools
• Long term view applies to standards
  • groundwork for FRBR was laid in 1998
  • as the basis for RDA cataloguing rules, became final in 2010, still pending review

Costs and ROI

• Cost of current practices is unknown
  • some costs are known but unmeasured
  • publicly funded service environment
• ROI is difficult to calculate
• Library-specific data formats require nice systems solutions
• Require technical staff with specific expertise in library data

Rights issues

• Some data cannot be published openly
• Rights ownership may be unmanageably complex
  • recent word of improvements at the OCLC, for example
• Rights have perceived value
  • may publish lossy information, e.g. dc:title for preferred and alternate title, losing detail required for proper citation

Library-Specific Formats

• Not easily shared outside of the library community

• Heavy use of text strings rather than identifiers (still less "linkable" URIs)

• Lack of best practices or standardisation for using RDF with library data

  • some data is expressed on experimental basis

• Lack of shared terminology

  statement, heading, authority control

Recommendations

• Identify candidate data sets for early exposure as linked data

  • determine ROI of current practice
  • determine cost and ROI of exposing as LD

• Evaluate migration strategies

• Foster a discussion about open data and rights

Education

• Facilitate an ethos of innovation
  • small scale independent R&D within individual library organisations
• Identify literacy in terms of metadata modelling generally and linked data specifically for different roles in the library
• Include metadata design in L&IS education
• Increase participation in linked-data standardisation efforts

Design, Identify and link

• Translate library data and standards into linked-data
• Develop best practices
• Design user stories and exemplar UIs
• Assign unique identifiers (URIs) for all significant things in library data
• Create URIs for the items in library datasets
• Create explicit links between library datasets and to other well-used datasets

Curation

• Apply library experience in curation and long-term preservation to linked data (and other) datasets

• Preserve linked data vocabularies

  • crucial for being able to interpret datasets
  • low-hanging fruit?

Questions?

The Incubator Group home page:

http://www.w3.org/2005/Incubator/lld/

This presentation:

http://eris.okfn.org/ww/2011/06/nls/

http://homepages.inf.ed.ac.uk/v1wwaite/2011/06/nls/