Linked Data Survey results 3–Why and what institutions are consuming (Updated)

 

OCLC Research conducted an international linked data survey for implementers between 7 July and 15 August 2014. This is the third post in the series reporting the results. 

The two main reasons why the 72 linked data projects/services described in the survey that consume linked data are: to enhance their own data by consuming linked data from other sources (37) and provide a richer experience for users (35). Other reasons, in descending order: more effective internal metadata management; to experiment with combining different types of data into a single triple store; heard about linked data and wanted to try it out using linked data sources; a wish for greater accuracy and scope in search results; to improve Search Engine Optimization (SEO); and to meet a grant requirement.

The ways projects/services are using linked data sources (in order of the most frequently cited):

• Enrich bibliographic metadata or descriptions
• Interlinking
• As a reference source and to
• Harmonize data from multiple sources
• Automate authority control
• Enrich an application
• Dataset discovery
• Auto-suggest

The linked data sources that are used the most:

• id.loc.gov – 30
• DBpedia – 25
• GeoNames – 25
• VIAF – 24

Here’s the alphabetical list of the sources used; those that include uses by FAST, VIAF, WorldCat.org and WorldCat.org Works are asterisked.

Source	# of Projects	FAST	VIAF	WorldCat.org	WorldCat.org Works
British National Bibliography	3			*
Canadian Subject Headings	2
DBpedia	25	*
Dewey Decimal Classification	5			*
DPLA	4
Europeana	5
FAST	11			*	*
GeoNames	25	*
Getty’s AAT	10
id.loc.gov	30	*	*	*	*
ISNI	5		*
ORCID	5
RDF Book Mashup	1
The European Library	5
VIAF	24	*		*	*
Wikidata	7
WorldCat.org	12		*	*	*
WorldCat.org Works	6		*	*	*
Other	20

The other linked data sources consumed include:

• Bibliothèque nationale de France’s data.bnf.fr, an aggregation of its catalogs and the Galica digital library.
• Deutsche National Bibliothek’s Linked Data Service
• GEMET, GEneral Multilingual Environmental Thesaurus
• Heritage Data’s SENESCHAL (Semantic ENrichment Enabling Sustainability of arCHAeological Links), a set of linked data vocabularies for cultural heritage
• HISCO, History of Work Information System
• Hispana, an aggregation of digital collections of archives, libraries and museums from Spanish digital repositories
• Lexvo for languages
• Logainm.ie, place names database of Ireland
• Nomisma.org, providing URIs for concepts unique to numismatics
• Pleiades Gazetter of Ancient Places, a community-built gazetteer and graph of ancient places in the Greek and Roman world.
• Rådata nå!, Norwegian name authority file, one of the first to be available as linked open data.
• United Nation’s Food and Agriculture Organization’s AGROVAC 

Asked whether there were other data sources the respondent wished were available as linked data but isn’t yet, respondents noted:

• More authority files or thesauri (requested by several) or multilingual subject vocabulary
• [U.S.] Federal agencies’ data
• Grant data
• Individual artworks and digital objects from archaeological or museum databases
• OpenStreetMap
• Researcher identifiers from smaller data stores

Barriers or challenges encountered in using linked data resources included:

• Size of RDF dumps; volatility of data formats of dumps; lack of availability of dumps; lack of authority control within the dumps; issues with level of specificity in terms of trying to match concepts.
• What is published to the Internet as Linked Data is not always reuseable…Linked data without context is almost useless.
• Many services present like Linked Data aren’t really Linked Data.
• It’s difficult to get other institutions to do their own harmonization between objects and concepts.
• Lots of handcrafting at the moment, not many off the shelf tools that are useful for visualisation.
• Mapping of vocabulary requires a lot of manual work.
• Matching, disambiguating and aligning source data and the linked data resources.
• Not all resources that we would like to use as linked data are represented as URIs. Semantics that can represent library bibliographic data are not established yet.
• It always requires time to understand how the data are structured before using it.
• Disambiguation of terms across different languages is difficult.
• DBpedia resources are not stable. URIs and structure for resource description would change.
• The creation of controlled vocabularies in SKOS seems less intuitive then we’d like.
• Service reliability has been a factor with some resources.
• Unstable endpoints, datasets not being updated.

[Originally posted 2014-09-01, updated 2014-09-04]

Coming next: Linked Data Survey results–Why and what institutions are publishing