There is no single commonly accepted definition of a knowledge graph.

Recent initiatives building geospatial knowledge graphs include:

• US: KnowWhereGraph https://knowwheregraph.org (https://www.nsf.gov/awardsearch/showAward?AWD_ID=2033521 and previously https://www.nsf.gov/awardsearch/showAward?AWD_ID=1936677)
  • Public repositories: https://github.com/KnowWhereGraph
  • Based on Ontotext GraphDB (proprietary with free version)
• NL: https://geonovum.github.io/NEN3610-Linkeddata/ review available standards, manuals:
  • Rules for developing ontologies in the Web Ontology Language (OWL)
  • INSPIRE - Guidelines for the RDF encoding of spatial data
  • W3C - Spatial Data on the Web Best Practices
  • Meaningfully connecting information with BP4mc2 practical experiences
  • Linked Data Proxy (LDProxy)
• GR: GeoQA project buidling YAGO2geo https://geoqa.di.uoa.gr
  • https://yago2geo.di.uoa.gr
  • Contains administrative data taken from official datasets of three countries: Greece, UK and Ireland. YAGO2geo also contains geospatial information for the every country extracted from GADM as well as other types of features, such as lakes and streams, from OpenStreetMap.
• The EU knowledge graph (https://linkedopendata.eu/wiki/The_EU_Knowledge_Graph), e.g. the entry for Austria (https://linkedopendata.eu/wiki/Item:Q16) includes:
  • coordinate location ... the country centroid
  • locator map image
  • geoshape ... GeoJSON https://commons.wikimedia.org/wiki/Data:Austria.map
  • coordinates of northernmost point
  • coordinates of easternmost point
  • area
  • located in time zone
  • part of
  • shares border with
  • highest point
  • water as percent of area
  • lowest point
  • coordinates of westernmost point
  • contains administrative territorial entity
  • coordinates of geographic center
  • NUTS code
  • GeoNames ID
  • OSM relation ID

RDF triples can simply be archived as CSV, e.g. in what is known as N-Triples format:

However there are other formats and guidelines:

RDF 1.1 Concepts and Abstract Syntax defines a data model. It does not define a specific format to encode RDF data. A number of RDF formats exist, serving different purposes.

• Recommendation REC/OWL/encoding/turtle ... INSPIRE RDF data should be encoded in Turtle.
• Permission PMS/OWL/encoding/other ... INSPIRE RDF data may be encoded in other RDF formats, such as N-Triples and XML.

For example, exports of the EU Knowledge Graph can be accessed under https://data.linkedopendata.eu/ as turtle .ttl files.

TTL (pronounced ‘turtle’) stands for “Terse RDF Triple Language” and is a file format used to express RDF data—a common alternative to N-Triples, JSON-LD, and XML in the RDF space. The format .ttl is a W3C standard that is described as a “general-purpose language for representing information in the web”. Representing RDF, .ttl files store facts as triples, and TriG is an extension of TTL that accepts quads (to utilise named graphs). (source: https://www.oxfordsemantic.tech/faqs/what-is-a-ttl-file)

In addition to SPARQL (a W3C standard declarative query language for querying graph data) there is also Cypher for Neo4j which was opened up through the openCypher project in October 2015.

Spatial support is available via GeoSPARQL and Cypher spatial functions, respectively.

Challenges for archiving knowledge graphs include:

• Many thesauri are not stable enough for long-term archiving
• Graph database exports, e.g., from Neo4J may contain more information than can be represented in RFD triples. However, the Neosemantics plugin enables the use of RDF in Neo4j (import and export of RDF, e.g. in turtle format).