1 of 41

Knowledge Graph Foundations

2 of 41

Outline

  • Introduction and motivation
    • History and areas of application
  • Knowledge Representation and Query Languages
    • RDF og RDFS
    • OWL
    • SPARQL
  • Ontologies and Vocabularies
    • Some Standard Vocabularies
    • schema.org
  • Data Lifting Standards
    • RDB2RDF
    • GRDDL

3 of 41

Introduction

​

“A knowledge graph consists of a set of interconnected typed entities and their attributes.”

Popularized by Google in 2012

Now also used by Facebook, LinkedIn and Microsoft

4 of 41

The Semantic Web (“the Web of meanings”)

​

​

History

Semantic web

RDF and OWL

Knowledge graph

Application

Open innovation

Market Intelligence

IBM Watson

5 of 41

IBM Watson

6 of 41

“A knowledge graph is a set of typed entities (with attributes) which relate to one another by typed relationships.

​

The types of entities and relationships are defined in schemas that are called ontologies. Such defined types are called vocabulary.

​

In this section, we will introduce �the standard for representing knowledge graphs (RDF), �two standards for defining ontologies (RDFS and OWL) �and the standard for querying knowledge graphs (SPARQL).”

7 of 41

How to read this book

8 of 41

Outline

  • Introduction and motivation
    • History and areas of application
  • Knowledge Representation and Query Languages
    • RDF og RDFS
    • OWL
    • SPARQL
  • Ontologies and Vocabularies
    • Some Standard Vocabularies
    • schema.org
  • Data Lifting Standards
    • RDB2RDF
    • GRDDL

9 of 41

Representing

Knowledge Graphs

10 of 41

RDF - Resource Description Framework

  • Originally created by W3C in 1999�
  • Modern standard to describe entities
    • person, homepage, company
    • Anything we can identify!�
  • structured -> machine understandable -> interoperability among applications exchanging information on the web

11 of 41

RDF Triples/Statements

​

12 of 41

RDF Triples/Statements Example 1

​

​

Kristian

Taco

likes

13 of 41

RDF Triples/Statements Example 2

​

​

14 of 41

Node Types

  • IRI
    • International Resource Identifier�
  • Literal
    • strings, numbers or dates
    • datatype
      • language if specific datatype�
  • Blank
    • “this is an existing thing”

​

15 of 41

Node Types (cont.)

IRI�Blank�Literal

IRI�Blank

IRI

16 of 41

Node Types Example

​

​

IRI

Literal

IRI

17 of 41

Prefixes

18 of 41

RDF Serialising

  • Several syntaxes for storing and exchanging RDF
    • Turtle
    • RDF/XML
    • RDFa
    • N-Triples
    • NQUADS
    • JSON-LD

19 of 41

RDF Serialising Example

@prefix dbpedia: <http://dbpedia.org/resource/>

@prefix dbpedia-owl: <http://dbpedia.org/ontology/>�

dbpedia:Bolivia dbpedia-owl:longName "Plurinational State of Bolivia" .

Example with Turtle:

20 of 41

Defining Ontologies

21 of 41

RDFS - Resource Description Framework Schema

​

  • Ontologies are schemas�
  • Simple schema language for RDF knowledge graphs�
  • RDF defines the structure of the data �
  • RDFS defines semantic relationships�

22 of 41

RDFS Example

​

org:Organization rdf:type rdfs:Class .

org:Start-up a rdfs:Class .

org:hasHomePage rdf:type rdfs:Property .

org:Start-up rdfs:subClassOf org:Organization .

​

23 of 41

OWL - Web Ontology Language

  • OWL2 W3C Recommendation Status�
  • “RDFS extension” �
  • Vocabulary - types of entities and relationships defined in ontologies�
  • Additional vocabulary defined by the OWL schema
    • Cardinality
    • Property chain
    • Self-restriction
    • Symmetric
    • Reflexive

24 of 41

OWL Example

Example from the Travel ontology:

:borders rdf:type owl:ObjectProperty ,� owl:SymmetricProperty ;� owl:propertyChainAxiom ( :hasBoundary� :boundaryOf� ) .

​

25 of 41

Querying Knowledge Graphs

26 of 41

SPARQL

  • Standard query language for RDF and OWL�
  • Turtle syntax
    • SQL like�
  • Based on pattern-matching mechanism

​

​

Mer om TURTLE og SPARQL

De fire "ledige" fikser dette

27 of 41

“A knowledge graph is a set of typed entities (with attributes) which relate to one another by typed relationships.

​

The types of entities and relationships are defined in schemas that are called ontologies. Such defined types are called vocabulary.

​

In this section, we will introduce �the standard for representing knowledge graphs (RDF), �two standards for defining ontologies (RDFS and OWL) �and the standard for querying knowledge graphs (SPARQL).”

28 of 41

Oversikt

  • Introduction and motivation
    • History and areas of application
  • Knowledge Representation and Query Languages
    • RDF and RDFS
    • OWL
    • SPARQL
  • Ontologies and Vocabularies
    • Some Standard Vocabularies
    • schema.org
  • Data Lifting Standards
    • RDB2RDF
    • GRDDL

29 of 41

Some standard

schemas and vocabularies

30 of 41

Standard Ontologies and Vocabularies

  • Organisation Ontology�
  • GoodRealtions�
  • Data Cube Vocabulary
  • Friend-of-a-Friend (FOAF)
  • schema.org

​

31 of 41

schema.org

  • Introduced by Google, Yahoo! and Bing in 2011�
  • Over 10 million sites�
  • Not intended to be a “global ontology”
    • Improve display of search results

32 of 41

Oversikt

  • Introduction and motivation
    • History and areas of application
  • Knowledge Representation and Query Languages
    • RDF and RDFS
    • OWL
    • SPARQL
  • Ontologies and Vocabularies
    • Some Standard Vocabularies
    • schema.org
  • Data Lifting Standards
    • RDB2RDF
    • GRDDL

33 of 41

Data Lifting Standards

34 of 41

Data Lifting Standards

35 of 41

Why? 👀

Legacy data

36 of 41

RDB2RDF

Relational database 2 Resource description framework

Make a choice

  • Direct mapping vs
  • using a mapping language

37 of 41

Tools

Morph-RDB https://github.com/oeg-upm/morph-rdb

D2RQ: http://d2rq.org/ (view only)

Virtuoso20: http://virtuoso.openlinksw. com/dataspace/doc/dav/wiki/Main/VirtR2RML (R2RML)

​

38 of 41

GRDDL

X(HT)ML -> RDF

Crawl webpages, discover rdf standards, use GRDDL to convert to your own knowledge graph

39 of 41

Knowledge graph vs Linked data

Linked data

  • many RDF datasets linked together

A knowledge graph is a structured dataset that is compatible with the RDF data model and has an (OWL) ontology as its schema.

  • high integrity and reliability

40 of 41

RDF vs Linked Data vs Knowledge Graph

Y= Yes, NN = Not Necessarily, N = No, L= Limited

41 of 41

Web search vs enterprise

Web

  • Small hierarchy
  • good enough
  • Schema.org, 700 types

Enterprise

  • More functionality than just search
  • can make use of more complex structures