DLF 2011 - What open source tools need to create/use Linked Data
We need infrastructure, low-level ways to discover/pull Linked Data from others, instead of someone selecting and hardcoding searching of single big external repositories.
Might need authority/rating-style filters. Could build on trust networks already in place. Should libraries maintain some kind of authority/rating system? How much mediation is needed? LOD-LAM summit had a discussion of linking the library "core record" concept to Linked Data. "Core record" type elements could be used as an evaluation tool (i.e. does this linked data contain the core elements?)
We need a better infrastructure for understanding the metadata vocabularies that are out there.
Need to be clear on when we're evaluating:
- quality of triples themselves
- properties/metadata vocabularies used by different communities
- controlled vocabularies used as the subjects of triples
Tie in to an infrastructure needed to mint URIs for sharing Linked Data. Standard library solution for this is registries. This makes sense for minting URIs for properties. But maybe not for minting URIs for other purposes.
Issues of URI persistence when content management systems change. Can sometimes still cope when the lowest-level identifier stays the same but base pattern of URI changes.
The few best practices in the SemWeb community are inconsistent, and none have a library bent. Can the library community help with this? Give requirements/expectations to our vendors?
Examples of systems that are layering in Linked Data
- HubZero as an example of layering in Linked Data into an existing system. RDF/a inside their HTML.
- Maybe the V/FRBR work at IU too - http://vfrbr.info (XML, RDF, maybe rest-ful style URLs)
- Chronicling America http://chroniclingamerica.loc.gov/ - puts metadata and links to alternative representations and ORE resource maps in HTML <link> tags (view source)
- search engines are starting to look for these types of patterns
- schema.org defines the microdata properties that search engines are looking for
"Linters" and "distillers" (eg http://linter.structured-data.org/) are tools that help you look at source code in a more human-readable way
Open Metadata Registry as a helpful tool.
- local library's subject list on their web site could be registered here, could reach a very broad audience