Skip to content

Home

Jump to bottom Edit New page
Tim L edited this page Aug 17, 2013 · 47 revisions

Linked Data is hard to make, and hard to use. Thanks to the many efforts of people around the world, a wealth of Linked Data is available for reuse. But, after seven years of development, the methods used to collect, organize, transform, update, and publish Linked Data remain immature and ad hoc. Despite the claims that interconnected data sources can lead to greater insights on a particular problem, a myriad of practical aspects make it difficult for consumers to performing many kinds of data analysis. Linked Data sources are hard to find, access, and understand; their publishers are hard to contact; and most corpora are not maintained, are out of date, and are not easily reproducible by third parties.

Too many tools, too much configuration. Although the diversity in tools and methods is a sign of an emerging and promising technology, it unfortunately also makes it difficult for broader communities to adopt Linked Data as a viable and compelling means to publish their data. Currently, too much expertise and manual effort is required to install and configure the many tools needed to implement an end-to-end Linked Data application. And, once an application has been built, it is currently costly for the same developer to replicate the same system. This ad hoc nature makes it difficult for developers to collaboratively build a system; it makes it difficult for third parties to inspect, learn from, reproduce, and extend existing solutions; and it makes it difficult for non-Linked Data experts to adopt Linked Data technology.

Prizms is a tool resulting from four years of Linked Data collection, organization, transformation, and publishing. Its components have been used to support dozens of Linked Data applications, and its design embodies the best practices that have emerged from those experiences as well as experiences suggested by the broader Linked Data community.

  • Prizms provides an organizational structure that enables a data scientist to collect many versions of the many datasets that are provided by many source organizations. This so-called "SDV" organization is organized hierarchically on the file system, and is also reflected in the URI space for the resulting Linked Data. The file system is version-controlled using git, allowing for change tracking, distributed collaboration, and reproducibility.
  • Prizms provides an organizational structure that facilitates the distinction among source data (that which we received from a publisher), our manual preparations of that source data for transformation into (or analysis of) Linked Data, the reproducible automatic output of scripts, and the resulting RDF package to publish. This organization of workflow artifacts is called the "conversion cockpit" because it is the fixed location that a curator should sit to focus on the best possible Linked Data representation of the given data.
  • Prizms captures provenance throughout each step to retrieve, prepare, transform, and publish Linked Data. Provenance is captured automatically and can be manually asserted using the available scripts. The Linked Data published by a Prizms node is traceable from the first URL retrieval to the loading of final SPARQL endpoint named graph.
  • Prizms generates a wealth of metadata using not only the best practice vocabularies (DCTerms, DCAT, VoID), but also our own extensions (SPO Balance, Between-The-Edges). This metadata facilitates third party discovery, access, and use of the Linked Data available from a Prizms node.

(coming up: an explanation for how Prizms leverages CKAN and LODSPeaKer capabilities)

Prizms

Presentations (newest on top):

Components

Prizms incorporates these independent Linked Data software:

  • csv2rdf4lod-automation provides automated, transparent, and reproducible RDF conversion
  • DataFAQs, a framework for LINKED DATA QUALITY REPORTS
  • vsr, Visual Strategies for RDF, brings Linked Data into the visual itself
  • LODSPeaKr Linked Data servering to humans and machines

Long Term Objective:

Add a custom footer
Add a custom sidebar

Clone this wiki locally