From property graphs to knowledge graphs

PhD Defense - Julian BRUYAT - 2024-06-03

1

Jury

• Catherine FARON, Reviewer
• Olaf HARTIG, Reviewer
• Anastasia DIMOU, Examinator
• Jose Emilio LABRA GAYO, Examinator
• Philippe LAMARRE, Examinator
• Frédérique LAFOREST, Advisor
• Pierre-Antoine CHAMPIN, Co-advisor
• Lionel MÉDINI, Co-advisor

​

​

​

Property graphs and RDF graphs: Knowledge graphs

2/39

Property graphs and RDF graphs: Property graphs

​

3/39

:Person

name: Tintin

job: Reporter

:Person

name: Haddock

:Animal

name: Snowy

:owns

:travelsWith

since: The crab with the golden claws

Vendors: Neo4j, JanusGraph, Amazon Neptune, Azure Cosmos DB…

Running example as a property graph

Property graphs and RDF graphs: Property graphs

​

​

4/39

Screenshot of the user interface of Neo4j

Property graphs and RDF graphs: RDF graphs

5/39

@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .

@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .

@prefix ex: <http://www.example.org/> .

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

@prefix foaf: <http://xmlns.com/foaf/0.1/> .

@prefix schema: <http://schema.org/> .

​

_:tintin rdf:type foaf:Person .

_:tintin foaf:name "Tintin" .

_:tintin schema:job dbpedia:Reporter .

​

dbpedia:Reporter schema:name "Reporter" .

​

_:snowy rdf:type dbpedia:Animal .

_:snowy dbpedia:name "Snowy" .

​

_:haddock rdf:type foaf:Person .

_:haddock foaf:name "Haddock" .

​

_:tintin ex:owns _:snowy .

​

_:travelers� ex:traveler _:tintin� ex:traveler _:haddock

ex:since "The crab with the golden claws" .

Running example as an RDF graph written in Turtle

Running example as an RDF graph in a graphical representation

Property graphs and RDF graphs: RDF-star graphs

6/39

@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .

@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .

@prefix ex: <http://www.example.org/> .

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

@prefix foaf: <http://xmlns.com/foaf/0.1/> .

@prefix schema: <http://schema.org/> .

​

_:tintin rdf:type foaf:Person .

_:tintin foaf:name "Tintin" .

_:tintin schema:job dbpedia:Reporter .

​

dbpedia:Reporter schema:name "Reporter" .

​

_:snowy rdf:type dbpedia:Animal .

_:snowy dbpedia:name "Snowy" .

​

_:haddock rdf:type foaf:Person .

_:haddock foaf:name "Haddock" .

​

_:tintin ex:owns _:snowy .

_:tintin ex:travelsWith _:haddock .

​

# RDF-star

<< _:tintin ex:travelsWith _:haddock >>

ex:since "The crab with the golden claws".

Running example as an RDF-star graph written in Turtle-star

Running example as an RDF-star graph in a graphical representation

Property graphs and RDF graphs: RDF graphs

​

• W3C standard (1999, 2004, 2014, 2025?)

​

• A lot of different standards developed around it
• Querying : SPARQL
• Reasoning: RDFS, OWL
• Data integration: R2RML, JSON-LD
• Validation: SHACL

7/39

Property graphs and RDF graphs

8/39

Even though they both rely on a graph data structure, property graphs and RDF graphs are very different

Objectives

9/39

Convert PG data to RDF to benefit from its advantages

in terms of reasoning, data integration, data validation…

conversion

reversibility?

State of the art

• NeoSemantics (Jesús Barrasa’s Neo4j plugin for PG - RDF interoperability)

10/39

:node1 rdf:type :Person .

:node1 :name "Tintin" .

​

:node1 :travelsWith :node2 .

<< :node1 :travelsWith :node2 >> :since "The crab …" .

​

:node2 rdf:type :Person .

Conversion operated by NeoSemantics, a Neo4j plugin for PG - RDF interoperability

:Person

name: Tintin

:Person

:travelsWith

since: The crab with the golden claws

​

• The Property Graph ontology [1]

[1] Tomaszuk, D., Angles, R., & Thakkar, H. (2020). Pgo: Describing property graphs in rdf. IEEE Access, 8, 118355-118369.

_:pg pgo:hasNode _:node1 .

_:pg pgo:hasNode _:node2 .

_:pg pgo:hasEdge _:edge3 .

​

_:node1 rdf:type pgo:Node .

_:node1 pgo:label "Person" .

_:node1 pgo:hasNodeProperty _:property1 .

_:property1 pgo:key "name" .

_:property1 pgo:value "Tintin" .

​

_:node2 rdf:type pgo:Node .

_:node2 pgo:label "Person" .

​

_:edge3 rdf:type pgo:Edge .

_:edge3 rdf:startNode _:node1 .

_:edge3 rdf:endNode _:node2 .

_:edge3 pgo:hasEdgeProperty _:property3 .

​

# ...

State of the art

• Hartig’s transformations [2,3]
• The user defines a configuration = functions that maps components of the PG (property keys, property values, labels and nodes/edges) to RDF terms
• The user chooses one approach among two for edge representation for all edges

RDF based approach

RDF-star based approach [2]

11/39

[2] Hartig, O. (2019, September). Foundations to Query Labeled Property Graphs using SPARQL. In SEM4TRA-AMAR@ SEMANTiCS.

[3] Khayatbashi, S., Ferrada, S., & Hartig, O. (2022, June). Converting property graphs to RDF: a preliminary study of the practical impact of different mappings. In Proceedings of the 5th ACM SIGMOD Joint International Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA) (pp. 1-9).

dbpedia:Tintin pgo:label "Person" .

dbpedia:Tintin foaf:name "Tintin" .

​

dbpedia:Haddock pgo:label "Person" .

​

dbpedia:Tintin :travelsWith dbpedia:Haddock .

​

<< dbpedia:Tintin :travelsWith dbpedia:Haddock >>

:since "The crab …" .

dbpedia:Tintin pgo:label "Person" .

dbpedia:Tintin foaf:name "Tintin" .

​

dbpedia:Haddock pgo:label "Person" .

​

ex:travelers ex:relation "Travellers" .

ex:travelers ex:traveler dbpedia:Tintin .

ex:travelers ex:traveler dbpedia:Haddock .

​

ex:travelers ex:since "The crab …"

:Person

name: Tintin

:Person

:travelsWith

since: The crab with the golden claws

​

State of the art

• All existing works propose a uniform conversion for all nodes, edges, properties
• However, this is not always the desired behavior

How to design a mapping language allowing users to produce idiomatic RDF graphs from PGs?

12/39

:Person

name: Tintin

job: Reporter

:Person

name: Haddock

:Animal

name: Snowy

:owns

:travelsWith

since: The crab with the golden claws

_:tintin ex:owns _:snowy .

​

_:travelers ex:traveler _:tintin .�_:travelers ex:traveler _:haddock .

_:travelers

ex:since "The crab with the golden claws" .

A part of the RDF graph presented earlier

Running example PG

Idiomatic RDF graph = a graph that uses common model structures

Motivation

How to design a mapping language allowing users to produce idiomatic RDF graphs from PGs?

13/39

* Under some conditions

Contributions

• A low level conversion: PREC-C
• A schema-based conversion: PRSC
• An auto-completion engine: Shacled Turtle

14/39

PREC-C & PRSC: https://github.com/BruJu/PREC

Shacled Turtle: https://github.com/BruJu/Shacled-Turtle

PREC-C: Example

15/39

:Person

name: Tintin

job: Reporter

:Person

name: Haddock

:Animal

name: Snowy

:owns

:travelsWith

since: The crab with� the golden claws

Property graph

Produced RDF graph

Context

(pseudo code)

Model edges as

?source ?edgeIRI ?destination��Model properties as� ?holder ?propertyIRI ?value

​

PREC = Property graph to RDF graph Experimental Converter�PREC-C = PREC-Context

:travelsWith → model it as

?self ex:traveler ?source

?self ex:traveler ?destination

:owns → ex:owns

since → ex:since

job → schema:job

but model it as

?holder ?propertyIRI ?self

?self schema:name ?value

name on a :Person → foaf:name

name → dbpedia:name

:Animal → dbpedia:Animal

:Person → foaf:Person

_:tintin rdf:type :Person .

_:haddock rdf:type :Person .

_:snowy rdf:type :Animal .

_:snowy :name "Snowy" .

_:tintin :name "Tintin" .

_:haddock :name "Haddock" .

_:tintin :job "Reporter" .

<< _:tintin :travelsWith _:haddock >>

:since "The crab with the golden claws" .

_:tintin :owns _:snowy .

_:tintin :travelsWith _:haddock .

_:tintin rdf:type foaf:Person .

_:haddock rdf:type foaf:Person .

_:snowy rdf:type dbpedia:Animal .

_:snowy dbpedia:name "Snowy" .

_:tintin dbpedia:name "Tintin" .

_:haddock dbpedia:name "Haddock" .

_:travelers ex:traveler _:tintin .�_:travelers ex:traveler _:haddock .

_:snowy dbpedia:name "Snowy" .

_:tintin foaf:name "Tintin" .      

_:haddock foaf:name "Haddock" .      

_:tintin schema:job _:reporter .

_:reporter schema:name "Reporter" .

_:tintin ex:owns _:snowy .

<< _:tintin :travelsWith _:haddock >>

ex:since "The crab with the golden claws" .

_:travelers ex:since "The crab with the golden claws" .

+

→

PREC-C: Contexts are RDF graphs

16/39

:Person → foaf:Person

​

​

:Animal → dbpedia:Animal

​

name → dbpedia:name

name on a :Person → foaf:name

​

job → schema:job

but model it as

?holder ?propertyIRI ?self

?self schema:name ?value

​

since → ex:since

​

:owns → ex:owns

​

​

:travelsWith → model it as

?self ex:traveler ?source

?self ex:traveler ?destination

​

Model properties as� ?holder ?propertyIRI ?propertyValue

​

Model edges as

?source ?edgeIRI ?destination

In the previous slide

In real life

​

prec:Properties prec:templatedBy prec:DirectTriples .

prec:Edges prec:templatedBy prec:RDFStarOccurrence .

​

[] a prec:NodeRule ;

prec:label "Person" ;

prec:nodeLabelIRI foaf:Person .

​

[] a prec:NodeRule ;

prec:label "Animal" ;

prec:nodeLabelIRI dbpedia:Animal .

​

[] a prec:PropertyRule ;

prec:propertyKey "name" ;

prec:propertyIRI dbpedia:name .

​

[] a prec:PropertyRule ;

prec:propertyKey "name" ;

prec:label "Person" ; prec:onKind prec:Node ;

prec:propertyIRI foaf:name .

​

[] a prec:PropertyRule ;

prec:propertyKey "job" ;

prec:propertyIRI schema:job ;

prec:templatedBy [

prec:produces << pvar:holder pvar:propertyIRI pvar:self >> ;

prec:produces << pvar:self schema:name pvar:value >> ;

] .

​

​

[] a prec:PropertyRule ;

prec:propertyKey "since" ;

prec:propertyIRI ex:since .

​

[] a prec:EdgeRule ;

prec:label "owns" ;

prec:edgeIRI ex:owns .

​

[] a prec:EdgeRule ;

prec:label "travelsWith" ;

prec:templatedBy [

prec:produces << pvar:self ex:traveler pvar:source >> ;

prec:produces << pvar:self ex:traveler pvar:destination >> ;

] .

Terms used in this example:

​

prec:Properties

prec:templatedBy

prec:DirectTriples

prec:Edges

prec:RDFStarOccurrence

prec:NodeRule

prec:label

prec:nodeLabelIRI

prec:PropertyRule

prec:propertyKey

prec:propertyIRI

prec:produces

prec:onKind

prec:EdgeRule

prec:edgeIRI

pvar:self

pvar:holder

pvar:propertyIRI

pvar:value

pvar:source

pvar:destination

​

[] a prec:PropertyRule ;

prec:propertyKey "job" ;

prec:propertyIRI schema:job ;

prec:templatedBy [

prec:produces << pvar:holder pvar:propertyIRI pvar:self >> ;

prec:produces << pvar:self schema:name pvar:value >> ;

] .

PREC-C: Design and issues

​

• Design:
• Converts all nodes, edges and properties
• Can customize how nodes, edges and properties are converted by assigning to rules a new template graph
• Issues:
• Big ontology with a lot of terms and concepts (fallback templates, substitutions…)
• The priority system is brittle

​

​

• Expected to be hard to use and hard to study (esp. reversibility)

17/39

:Person :Animal

name: Enid

race: Werewolf

name on a :Person = foaf:name

name on an :Animal = schema:name

Which one do we pick?

Motivation

How to design a mapping language allowing users to produce idiomatic RDF graphs from PGs?

18/39

* Under some conditions

Contributions

• A low level conversion: PREC-C
• A schema-based conversion: PRSC
• An auto-completion engine: Shacled Turtle

19

PREC-C & PRSC: https://github.com/BruJu/PREC

Shacled Turtle: https://github.com/BruJu/Shacled-Turtle

PRSC: a schema-based conversion

20/39

:Person

name: Tintin

:Person

name: Tintin

job: Reporter

:Person

name: Haddock

:Animal

name: Snowy

:owns

:travelsWith

since: The crab with� the golden claws

:Person

name:

:Animal

name:

:owns

:travelsWith

since:

Running example PG

Schema

PRSC = Property graph to RDF graph Schema-based Converter

You will struggle to pronounce it until you realize you can pronounce it “presc”

:Person

name:

job:

PRSC: The context rises from the schema

21/39

:Person

name:

:Animal

name:

:owns

:travelsWith

since:

Schema

​

:PersonRule a prec:PRSCNodeRule ;

prec:label "Person" ;

prec:propertyKey "name" ;

prec:produces

​

​

​

​

:AnimalRule a prec:PRSCNodeRule ;

prec:label "Animal" ;

prec:propertyKey "name" ;

prec:produces

​

​

​

:OwnsRule a prec:PRSCEdgeRule ;

prec:label "owns" ;

prec:produces�

​

:TravelsWithRule a prec:PRSCEdgeRule ;

prec:label "travelsWith" ;

prec:propertyKey "since" ;

prec:produces

​

​

​

​

​

​

​

​

​

<< pvar:self rdf:type foaf:Person >> ,

<< pvar:self foaf:name "name"^^prec:valueOf >> .

​

​

​

​

​

​

<< pvar:self rdf:type dbpedia:Animal >> ,

<< pvar:self schema:name "name"^^prec:valueOf >> .

​

​

​

� << pvar:source ex:owns pvar:destination >> .

​

​

​

​

​

<< pvar:self ex:traveler pvar:source >> ,

<< pvar:self ex:traveler pvar:destination >> ,

<< pvar:self ex:since "since"^^prec:valueOf >> .

PRSC context

PRSC: The output

22/39

:Person

name: Tintin

:Person

name: Haddock

:Animal

name: Snowy

:owns

:travelsWith

since: The crab with� the golden claws

Running example PG

​

:PersonRule a prec:PRSCNodeRule ;

prec:label "Person" ;

prec:propertyKey "name" ;

prec:produces

<< pvar:self rdf:type foaf:Person >> ,

<< pvar:self foaf:name "name"^^prec:valueOf >> .

​

​

:AnimalRule a prec:PRSCNodeRule ;

prec:label "Animal" ;

prec:propertyKey "name" ;

prec:produces

<< pvar:self rdf:type dbpedia:Animal >> ,

<< pvar:self schema:name "name"^^prec:valueOf >> .

​

:OwnsRule a prec:PRSCEdgeRule ;

prec:label "owns" ;

prec:produces

<< pvar:source ex:owns pvar:destination >> .

​

:TravelsWithRule a prec:PRSCEdgeRule ;

prec:label "travelsWith" ;

prec:propertyKey "since" ;

prec:produces

<< pvar:self ex:traveler pvar:source >> ,

<< pvar:self ex:traveler pvar:destination >> ,

<< pvar:self ex:since "since"^^prec:valueOf >> .

​

​

​

​

​

_:tintin rdf:type foaf:Person .

_:tintin foaf:name "Tintin" .

​

_:haddock rdf:type foaf:Person .

_:haddock foaf:name "Haddock" .

​

​

​

_:snowy rdf:type dbpedia:Animal .

_:snowy schema:name "Snowy" .

​

​

​

​

_:tintin ex:owns _:snowy .

​

​

​

​

​

_:travelers ex:traveler _:tintin .

_:travelers ex:traveler _:haddock .

_:travelers ex:since

"The crab with the golden claws" .

PRSC context

Produced RDF graph

PRSC: Reversibility

• Is the conversion operated by PRSC reversible = does it lose any information? It depends on the context!

​

• PRSC well-behaved contexts
• Proved to be reversible
• Three criteria:
• “No information loss”
• “Signature template triple”
• “Element provenance”

23/39

Well-behaved

contexts

Reversible contexts

All PRSC contexts

PRSC: Well-behaved contexts - No information loss

• Inside each template graph, each property of the type must have a discriminative template triple.
• For edges, the source (resp. the destination) must also have a discriminative template triple.

24/39

:Person

givenName: Archibald

familyName: Haddock

:PersonRule

prec:produces

<< pvar:self foaf:name "givenName"^^prec:valueOf >> .

✘

_:haddock foaf:name "Archibald" .

:PersonRule

prec:produces

<< pvar:self foaf:name "givenName"^^prec:valueOf >> ,

<< pvar:self foaf:name "familyName"^^prec:valueOf >> .

✘

_:haddock foaf:name "Archibald" .

_:haddock foaf:name "Haddock" .

:PersonRule

prec:produces

<< pvar:self foaf:givenName "givenName"^^prec:valueOf >> ,

<< pvar:self foaf:familyName "familyName"^^prec:valueOf >> .

_:haddock foaf:givenName "Archibald" .

_:haddock foaf:familyName "Haddock" .

✔

PRSC: Well-behaved contexts - Signature template triple

• Each template graph must have a template triple that no other template graph has

25/39

:Person

name: Tintin

:Animal

name: Snowy

:PersonRule

prec:produces

<< pvar:self foaf:name "name"^^prec:valueOf >> .

​

​

:AnimalRule

prec:produces

<< pvar:self foaf:name "name"^^prec:valueOf >> .

✘

_:tintin foaf:name "Tintin" .

_:snowy foaf:name "Snowy" .

:PersonRule

prec:produces� << pvar:self rdf:type foaf:Person >> ,

<< pvar:self foaf:name "name"^^prec:valueOf >> .

​

​

:AnimalRule

prec:produces� << pvar:self rdf:type dbpedia:Animal >> ,

<< pvar:self foaf:name "name"^^prec:valueOf >> .

_:tintin rdf:type foaf:Person .

_:tintin foaf:name "Tintin" .

​

_:snowy rdf:type dbpedia:Animal .

_:snowy foaf:name "Snowy" .

✔

:PersonRule

prec:produces

<< pvar:self foaf:name "name"^^prec:valueOf >> .

​

​

:AnimalRule

prec:produces

<< pvar:self schema:name "name"^^prec:valueOf >> .

_:tintin foaf:name "Tintin" .

_:snowy schema:name "Snowy" .

✔

PRSC: Well-behaved contexts - Element provenance

• All template triples in all template graphs must have pvar:self

26/39

:Person

name: Tintin

:Person

name: Haddock

:travelsWith

:travelsWith

:TravelsWithRule a prec:PRSCEdgeRule ;

prec:label "travelsWith" ;

prec:produces

<< pvar:source ex:travelsWith pvar:destination >> .

_:tintin ex:travelsWith _:haddock .

“RDF graphs are defined as sets of triples”

✘

:TravelsWithRule a prec:PRSCEdgeRule ;

prec:label "travelsWith" ;

prec:produces

<< pvar:self ex:traveler pvar:source >> ,

<< pvar:self ex:traveler pvar:destination >> .

✔

_:travel1 ex:traveler _:tintin .

_:travel1 ex:traveler _:haddock .

​

_:travel2 ex:traveler _:tintin .

_:travel2 ex:traveler _:haddock .

PRSC: Reversibility algorithm for well-behaved contexts

1. Identify the PG type of each RDF blank node
• Element provenance / Signature template triple
• See Theorem 3
2. Identify the RDF triples generated for each PG node and edge
• Element provenance / Signature template triple
• See Theorem 4
3. Populate the properties of each PG node and edge
• No information loss
• See Theorem 5

27/39

PRSC: Edge-unique types

28/39

• Element provenance: “All template triples in all template graphs must have pvar:self”
• With this criterion, all PG edges are converted to an RDF node
• Edge-unique types: some edge types are meant to be used at most once between two given nodes.
• Those edges could be represented as single RDF triples without loss of information

:Person

name: Tintin

:Animal

name: Snowy

:owns

:Animal

name: The llama in Prisoners of the sun

PRSC: Edge-unique types

29/39

• Edge-unique types support: Some edge types may not comply with the “element provenance” criterion if all template triples used for edge unique types comply with the following rules:
• The source and the destination must be in all template triples.
• Each triple of the template graph must be distinguishable from all other template triples in the whole context

:Person

name: Tintin

:Animal

name: Snowy

:owns

:Animal

name: The llama in “Prisoners of the sun”

:OwnsEdgeRule a prec:PRSCEdgeRule ;

prec:produces

<< pvar:source ex:owns pvar:destination >> .

_:tintin ex:owns _:snowy .

✔

-> No other template triple is allowed to use ex:owns in predicate position.

PROVED BY THEOREM 7

PRSC: Design and issues

• Design:
• Simple system based on schemas
• Some contexts are proved to be reversible:
• Well-behaved contexts
• Edge-unique extension

​

• Issues:
• A given context can only convert a given subset of PGs
• The schema language is a bit strict: no meta properties, no optional properties, closed types

30/39

Motivation

How to design a mapping language allowing users to produce idiomatic RDF graphs from PGs?

31/39

* Under some conditions

Contributions

• A low level conversion: PREC-C
• A schema based conversion: PRSC
• An auto-completion engine: Shacled Turtle

32/39

PREC-C & PRSC: https://github.com/BruJu/PREC

Shacled Turtle: https://github.com/BruJu/Shacled-Turtle

Shacled Turtle: What do I write in my context?

• The PREC ontology is composed of many terms.
• How to help the user writing contexts?

33/39

Terms used in this presentation:

​

prec:Properties

prec:templatedBy

prec:DirectTriples

prec:Edges

prec:RDFStarOccurrence

prec:NodeRule

prec:label

prec:nodeLabelIRI

prec:PropertyRule

prec:propertyKey

prec:propertyIRI

prec:onKind

prec:EdgeRule

prec:edgeIRI

pvar:self

pvar:holder

pvar:propertyIRI

pvar:value

pvar:source

pvar:destination

prec:PRSCNodeRule

prec:PRSCEdgeRule

prec:produces

​

​

Shacled Turtle: Principle

34/39

Shacled Turtle: How to generate the help?

• Schemas (RDFS, OWL, SHACL…) describe the expected structure of the graph.
• In this example: PRSCNodeRules have rdf:type, prec:label, prec:propertyKey and prec:produces.

​

• Schemas may also provide textual information: rdfs:label, sh:name, rdfs:comment, sh:description.

​

• A generic autocompletion engine, leveraging schemas, could be useful beyond PREC

35/39

Shacled Turtle: Experiments

• Exercice: convert a text into Turtle using the autocompletion engine
• Two ontologies: schema.org (hundreds of terms) and foaf (a few dozen of terms, strong inference system)
• ~30 volunteers
• Results: disappointing in both cases
• Users felt restricted -> promote relevant terms instead of filtering irrelevant ones
• Users want to understand why the engine does some choices -> use Shacled Turtle analysis to improve the descriptions

​

• In the context of writing an RML or a PREC mapping (closed ontologies): left for future works

36/39

Conclusion

37/39

Design a mapping language allowing users to produce idiomatic RDF graphs from PGs

​

Conclusion

38/39

Design a mapping language allowing users to produce idiomatic RDF graphs from PGs

​

Perspectives

PREC-C / PRSC

• Improve PRSC with PREC-C properties
• Run a user evaluation
• Reuse PREC templating system in other transformation systems

​

Shacled Turtle

• Improve the tool
• Experiment on closed ontologies, in particular PREC

39/39

The characterization function κ

40/39

Well-behaved context: Formal definition

41/39

Shacled Turtle - Architecture

42/39

This slide was not used during the defense

Shacled Turtle - Example

43/39

- The range of founder is ex:Person

- ex:PersonShape targets the ex:Person class

- ex:name is a property of ex:PersonShape shape

This slide was not used during the defense

Substitution predicates

44/39

[] a prec:PropertyRule ;

prec:propertyKey "job" ;

prec:propertyIRI schema:job ;

prec:templatedBy [

prec:produces << pvar:holder pvar:propertyIRI pvar:self >> ;

prec:produces << pvar:self schema:name pvar:value >> ;

] .

​

prec:propertyIRI a prec:SubstitutionPredicate

prec:substitutionTarget pvar:propertyIRI .

[] a prec:PropertyRule ;

prec:propertyKey "job" ;

prec:substitute [

prec:from pvar:propertyIRI

prec:to schema:job

] ;

prec:templatedBy [

prec:produces << pvar:holder pvar:propertyIRI pvar:self >> ;

prec:produces << pvar:self schema:name pvar:value >> ;

] .

​

This slide was not used during the defense