Linked Data

Winter 2019
Instructor: Raghava Mutharaju
IIIT Delhi


  • Individual data silos
  • Intra and inter connected/linked data
  • Links are typed: relation between two entities can be indicated
  • Huge amount of data is generated continuously
    • How can data be reused easily?
    • How can we enable discovery of relevant data?
    • How can we enable applications to integrate data from multiple sources?
  • Linked Data helps in discovering, accessing, integrating, and using data
  • WWW helps in connecting and consuming documents. Through Linked Data, it can also help in connecting and consuming data
  • Rationale for Linked Data
    • Structured data enables sophisticated processing
    • Hyperlinks connect distributed data
      • Makes the data discoverable
    • From data islands to a global data space
      • RDF links things, not just documents
      • RDF links are typed
      • Web of Data (analogous to Web of documents)

Linked Open Data Cloud

Principles of Linked Data

  • Linked Data refers to a set of best practices/principles for publishing and interlinking structured data on the Web
  • These principles were introduced by Tim Berners-Lee
  • Linked Data principles
    • Use URIs as names for things
    • Use HTTP URIs, so that people can look up those names (make URIs dereferenceable)
    • When someone looks up a URI, provide useful information, using the standards (RDF, SPARQL)
    • Include links to other URIs, so that they can discover more things

Naming Things using URIs

  • Resources (real-world entities, abstract concepts) need to be uniquely identified
  • HTTP URIs is a good mechanism to identify resources
    • They provide a simple mechanism to create globally unique identifiers
    • They provide a means of accessing the identified entity

Make URIs Defererenceable

  • HTTP clients can look up the URI and retrieve the resource's description
  • Content negotiation
    • HTTP clients can include in their headers what type of description they are looking for
    • HTML description is for human consumption
    • RDF description is for machine consumption
    • Two strategies can be used make URIs: 303 URIs and Hash URIs

Make URIs Defererenceable

  • 303 URIs
    1. Client performs a HTTP GET on a URI. It indicates either Accept: application/rdf+xml or Accept: text/html in the header
    2. Server responds with a HTTP 303 See Other response that includes the URI of the resource in requested format
    3. Client performs HTTP GET on this URI
    4. Server responds with the document and 200 OK

Make URIs Defererenceable

  • Hash URIs
    1. 303 URIs requires two HTTP requests. Hash URIs avoids that.
    2. URIs can have a base part and a fragment identifier (after #)
    3. Client removes the fragment identifier (HTTP protocol requirement) and sends rest of the URI to the server
    4. Server returns the details of all the fragment identifiers
    5. So client might end up with large number of triples that it is not interested in

Provide Useful RDF Information

  • When publishing Linked Data on the Web, data is represented in the form of RDF
  • The two most commonly used serialization formats when publishing Linked Data are rdf/xml and RDFa
  • RDF features that should be best avoided when publishing Linked Data
    • RDF reification: querying becomes cumbersome
    • RDF collections: querying becomes cumbersome
    • Blank nodes: they have local scope

Include Links to Other Things

  • Include external RDF links to other data sources so that data islands can be transformed into global, interconnected data space
  • There are three important types of RDF links
    1. Relationship Links: they point to related things in other data sources
    2. Identity Links: they point to URI aliases
    3. Vocabulary Links: they point to the definition of vocabulary terms that are used to represent the data


  • A unifying data model: RDF
  • A standardized data access mechanism: HTTP
  • Hyperlink-based data discovery: URIs
  • Self-descriptive data: shared vocabularies and data descriptions

5 Star Rating Scheme

  1. One Star: data is available on the web (whatever format), but with an open license
  2. Two Stars: data is available as machine-readable structured data (e.g., Microsoft Excel instead of a scanned image of a table)
  3. Three Stars: data is available as (2) but in a non-proprietary format (e.g., CSV instead of Excel)
  4. Four Stars: data is available according to all the above, plus the use of open standards from the W3C (RDF and SPARQL) to identify things, so that people can link to it
  5. Five Stars: data is available according to all the above, plus outgoing links to other people’s data to provide context

Publishing Data about Data

  • Two common mechanisms to describe data are Semantic Sitemaps and voiD descriptions
    • Semantic Sitemaps: similar to sitemap (url,loc,lastmod,changefreq), but has more elements such as location of data dumps, SPARQL endpoints
    • voiD: stands for Vocabulary of Interlinked Data. It is similar to Semantic Sitemap, except that it is in RDF. Subsets of a dataset can also be defined and linked to other datasets

Using Vocabularies to Describe Data

  • There are several popular existing vocabularies that can be reused to describe data, rather than reinventing the terms
  • If existing vocabularies are used then there is a higher chance that your data will be consumed/understood by applications
    • Dublin Core: defines general metadata attributes such as title, creator, date and subject
    • Friend-of-a-Friend (FOAF): defines terms for describing persons, their activities and their relations to other people and objects
    • created by major search engines to discover data for search
    • SKOS: bridge between different knowledge organization systems (thesaurus, taxonomy) to Semantic Web standards (RDF, OWL)
  • Basic Geo (WGS84): defines terms such as lat and long for describing geographically-located things
  • Music Ontology: defines terms for describing various aspects related to music, such as artists, albums, tracks, performances and arrangements
  • Bibliographic Ontology (BIBO): provides concepts and properties for describing citations and bibliographic references (i.e., quotes, books, articles, etc.)

  • Linked Open Vocabularies (LOV): searchable collection of several vocabularies

Linked Data Publishing Patterns

Linked Data Publishing Patterns
Image source:


  1. Linked Data: Evolving the Web into a Global Data Space. Tom Heath, Christian Bizer. 2011. pp 1-136. Morgan & Claypool.