Index

Knowledge Checks

Every check and exercise, each on its own page. 22 so far.

Knowledge Graph Foundations

Weekly self-scored checks, about five minutes each, drawn from that week's lecture materials.

Week 1

Forms, Sets & Worlds

Introduction to Conceptual Modelling

  • Plato's Idea Theory
  • set theory and hierarchy
  • classes, instances, properties
  • TBox / ABox / CBox
  • open and closed world

Self-scored quiz · 12 questions · ~5 min

Week 2

Control the Vocabulary

Controlled Vocabularies and the CBox

  • competency questions
  • vocabulary control define, reconcile, disambiguate
  • flat list, glossary, taxonomy, thesaurus
  • the CBox as a bridge
  • Ranganathan's canons and MECE
  • the coverage model

Self-scored quiz · 12 questions · ~5 min

Week 3

Concepts, Schemes & Triples

URIs, Taxonomies and SKOS

  • URIs and namespace prefixes
  • what a taxonomy needs
  • SKOS as data model and ontology
  • concept schemes, broader, topConceptOf, exactMatch
  • documentation properties
  • authority sources and the RDF triple

Self-scored quiz · 12 questions · ~5 min

Week 4

Ontology Foundations

Competency questions, design heuristics, RDF/RDFS/OWL

  • analysing a competency question time, instances, genre concept, geolocation
  • the three design heuristics rigid classes, sub classes that inherit and extend, classes drive properties
  • finding the TBox in your SKOS taxonomy
  • rdf:type, rdfs:subClassOf, owl:Class, owl:ObjectProperty, owl:DatatypeProperty
  • rdfs:domain and rdfs:range, rdfs:label and rdfs:comment
  • owl:Ontology and rdfs:isDefinedBy; Turtle semicolons and commas

Self-scored quiz · 12 questions · ~5 min

Module 5 tutorial

SPARQL Practice Room

Ten lessons, a task each, and a query window over a music graph in your browser

  • triples and the catch-all pattern; PREFIX, SELECT, WHERE from the module deck
  • FILTER, OPTIONAL and the open world, BIND and FILTER NOT EXISTS, ORDER BY, LIMIT, DISTINCT
  • COUNT and GROUP BY, property paths for query expansion, ASK, CONSTRUCT, DESCRIBE
  • upload your own Turtle and query it, or both graphs at once, with GRAPH ?g
  • final answer one of your competency questions in SPARQL and export it
  • every query runs in the browser (Oxigraph in WebAssembly); nothing is sent anywhere

Interactive tutorial · 10 lessons · 32 points · ~60 min

Module 5 tool

ABox Workbench

Add the facts: instances typed by your classes, controlled values picked from your taxonomy, questions answered live

  • three layers, three jobs the TBox gives the types, the SKOS taxonomy gives controlled values, the ABox is where the facts go; a concept is a value, never a class
  • Foundation load the module's one .ttl file (ontology, concept schemes, instances) or the ontology and the taxonomy as two files, read-only; a concept named like a class is flagged (Activity 1 happens in the files)
  • Activity 2, Build pick a class, name the instance, fill in its properties: a typed value, a link to another instance, or a controlled value from the taxonomy picker; every fact in words next to its Turtle; domain, range and datatype checked as you type, with the inference a reasoner would draw
  • a live graph of the facts, redrawn with every edit: instances, controlled values, classes and values as nodes, each fact as a labelled edge, a flash on what was just written; the answer graph lights the instances that answer a question and draws the skos:broader climb
  • questions as fill-in-the-blanks which class, and what must be true about it; subclass and skos:broader expansion; live answers and the first statement with no match; a detailed editor for two-hop questions
  • Activity 3 the SPARQL for each question in the deck's style, copied with the whole graph into the SPARQL Playground
  • review coverage per question, model use (classes with instances, properties used, controlled values in use), warrant per instance, findings in plain words with a link to fix each; exports the submission .ttl, the queries .txt, a report and the project
  • everything in the browser, autosaved to the tab; worked example: the module's music events graph with one question left blank to fill in

Design tool · 4 tabs · live graph · worked example · ~60 min

Modules 3 to 5 explainer

Two Jobs for SKOS

SKOS labels and notes on your classes; SKOS concepts as controlled values beside them

  • two kinds of SKOS property labels and notes are OWL annotation properties with no domain; broader, narrower, related and the mapping properties have domain and range skos:Concept
  • job one, annotations prefLabel, altLabel, hiddenLabel, definition, scopeNote, example, historyNote and changeNote on a class or a property, one preferred label per language, nothing changes for a reasoner
  • job two, controlled values rdf:type for the class, a property with range skos:Concept for the concept; rdfs:subClassOf propagates types, skos:broader is climbed by a query path
  • one word, two layers when a word earns a class and when a concept; rdfs:seeAlso to link them, never owl:sameAs, skos:exactMatch or rdf:type; punning left for later
  • six checks that keep the layers apart in any file, and which ones the ABox Workbench runs
  • twelve questions with explanations and links back to the section that teaches each one

Read-along explainer · 5 sections · 12-question check · ~20 min

Modules 1 to 5 explainer

Why the TBox Stays Small

ABox and TBox, class explosion, and why object properties, data properties and subclasses need care

  • the TBox lists kinds of thing, the ABox lists things, the CBox lists allowed values; two questions sort any statement; three layers, three speeds of change
  • class explosion a value in a class name multiplies the TBox by every dimension; an interactive counter for genres, instruments and cities; the adjective test
  • when a subclass is earned a new property or rule, the all-members-always promise, and members that keep landing in two siblings
  • object properties domain and range are inferences, a property per class or per value is the same explosion, sub-properties for real extra meaning, keep the domain as wide as the uses
  • data properties literals end the graph, city as a string versus a node, one property per datatype, no property named after another class, type every literal
  • separation of concerns as three files with three owners; seven mistakes to find in any Turtle file; ten questions with explanations

Read-along explainer · 6 sections · explosion counter · 10-question check · ~20 min

Modules 3 to 5 explainer

Requires, Two Ways

"An electric guitar requires an amplifier", modelled with SKOS concepts as values and with no SKOS at all

  • the sentence is a rule about every electric guitar and a fact about one guitar; the fact is the same object property in both models
  • with SKOS as values types are concepts, a thing says its type through a property, the rule is an Accessory Requirement individual with two values in the ABox
  • without SKOS types are classes, a thing says its type with rdf:type, the rule is an owl:someValuesFrom restriction on the class in the TBox
  • what a reasoner adds in each model (a property chain; the superclasses) and why neither flags a missing cable under the open world assumption
  • the gap query in both models; eight rows that decide between the models; twelve questions, six of them sorting Turtle into the model it belongs to

Read-along explainer · 6 sections · 12-question check · ~20 min

Modules 3 to 5 tool

Requires Workbench

Write the rules, add the gear, switch between SKOS-as-values and pure OWL, and watch the three layers rebuild

  • one switch with SKOS as values, or without SKOS; the Turtle for TBox, CBox and ABox regenerates and the rules are highlighted where they land
  • rules as rows (every instrument of this type needs an accessory of that type), inherited down the hierarchy through skos:broader or rdfs:subClassOf
  • gear with types and accessories; add a new instrument type and see one concept added, or one class and a schema-edit count
  • two queries answered live from what was written: what a type requires, and which instrument is missing what
  • what a reasoner adds in each model, and the open world note when a required accessory is missing
  • copy all three layers as one Turtle file to run the queries in the SPARQL Practice Room

Interactive tool · one switch · live Turtle and queries · ~20 min

Modules 3 to 5 explainer

Instruments in Three Layers

SKOS kinds and accessories, an OWL TBox of musicians and physical instruments, and the facts about one concert: built up until one question needs all three

  • kinds as a SKOS scheme twelve instrument kinds in three branches, labels, definitions, scope notes; a kind is a value, never a class
  • a kind is a concept, the instrument in your hands is an instance: five classes and ten properties, two of them with range skos:Concept (plays, is of kind), the class named Physical instrument on purpose
  • required accessories between kinds ex:requires as a sub-property of skos:related (the ISO 25964 way to add a relationship type), the symmetric inference, the S27 rule, inheritance through skos:broader* in the query
  • the ABox three musicians, three physical instruments, three makers, one concert; every line a type, a controlled value, a link or a value
  • one question, three layers the gear-list query line by line, its nine answers, and two more queries; the whole graph as a downloadable Turtle file for the Workbench, the Practice Room and the Playground
  • where everything in the music world goes a decision table from Bass guitar to Höfner to Rock to a music lesson
  • the same word in three layers labels may repeat, IRIs must differ; a namespace per layer, a naming convention, different labels, or punning (not in this course); the ex:musician clash between this file and the Module 5 graph
  • fourteen questions with explanations and links back to the section that teaches each one

Read-along explainer · 7 sections · worked data file · 14-question check · ~30 min

Module 1 exercise

Guess the Furniture

Live twenty questions, then draw the sets

  • students write and answer the questions live
  • yes/no, multiple choice, open-ended
  • big reveal when someone guesses
  • bonus round against the computer
  • sets and subsets drawn from the answers
  • each round exports to Excel, CSV, Markdown, JSON

Live classroom game · one file per round · then draw the sets · ~60 min

Module 1 check

What Goes Where

TBox, ABox and CBox in the music domain

  • sort music statements into the three boxes, Bad Bunny included
  • which box a questionnaire question reveals
  • flat list, glossary, taxonomy, thesaurus
  • synonyms, preferred terms, MECE
  • instance checking, realization, consistency, subsumption
  • why keep schema and data apart

Self-scored sorting game · 5 rounds · 32 points · ~8 min

Module 2 tool

Coverage Model

Choose your subgraph, then curate: coverage, completeness, disambiguation

  • four subgraphs genres and works, events and venues, instruments, musicians and works
  • use case, coverage targets, existing vocabulary per subgraph
  • 41 to 44 harvested terms from five sources with provenance and frequency
  • exact match, cluster, new candidate, documented exclusion
  • genus-differentia definitions and scope notes for homonyms
  • coverage graph by source and area, checked against a key
  • exports the reconciliation workbook and a SKOS starter

Curation exercise · 4 steps · answer key · ~90 min

Module 3 tutorial

SKOS in One Sitting

Labels, hierarchy and documentation

  • prefLabel, altLabel, hiddenLabel and the one-per-language rule
  • broader, narrower, related, top concepts, no cycles
  • definition, scopeNote, example, historyNote, editorialNote, changeNote
  • fill in the Turtle tap each value into its blank
  • a complete valid scheme to download

Read-along tutorial · 4 tap-to-place exercises · 33 points · ~25 min

Module 3 exercise

SKOS Crosswalk

Hand in your scheme, read the class's, map across subgraphs

  • one link per class (?class=class-a, ?class=class-b): separate folders, saved work and graph-store space; samples step aside for real data; Start over clears a device
  • Start screen per-team to-do list that ticks itself off, and a board showing which teams and students have uploaded
  • submit a SKOS Turtle file parsed and checked against S13, S14, S27, hierarchy loops, missing definitions
  • browse the four schemes one at a time or all at once: hierarchy graph side by side, labels, definitions, search across all
  • the class graph every concept a bubble, every relationship a line, force layout with drag, pan and zoom
  • map concepts across schemes with exactMatch, closeMatch, broadMatch, narrowMatch, relatedMatch
  • leads from shared labels, with planted false friends
  • mapping graph, transitivity and conflation checks, where the class disagrees
  • exports a mapping file in Turtle and a CSV
  • optional live backend Apache Jena Fuseki class graph store behind a small submit-and-list API

Class exercise · 5 steps · shared submissions · ~60 min

Module 3 game

SKOS in the Wild

The history and real-world use of SKOS, in seven levels

  • origins Desire, the 2004 Core Guide, the 2009 Recommendation, ISO 25964, MeSH in RDF (timeline to order)
  • the ilovemusic instruments and musicians taxonomy: fill in the Violin, place hidden, alt, related, scope note, broader
  • world tour UNESCO, Libris/DNB/BnF, STW/TheSoz, AGROVOC, EuroVoc/ADMS, ESCO, XKOS, MeSH; match each need to uneskos:contains, SKOS-XL, skos:notation, narrowMatch, xkos:Correspondence, xkos:maxLength, meshv pairs
  • SKOS vs OWL sort statements; mapping strength owl:sameAs / exactMatch / closeMatch with the Phytoplankton anti-pattern; ISO 25964 sub-properties
  • SKOS-XL the Stephen King / Richard Bachman reified labels in Turtle
  • query expansion live demo of skos:narrowerTransitive* over a mini library
  • quality patrol find omitted top concepts, a cycle, a disjoint-label violation and an orphan in a scheme
  • points, seven badges, a field rank and an exportable score card

Gamified tutorial · 7 levels · 64 points · 7 badges · ~40 min

Exercise

The CQ Forge

Competency Questions

  • strong vs. weak CQs and the five checks
  • the four shapes lookup, traversal, aggregation, inference
  • decomposing a CQ into classes, properties, controlled values, rules

Interactive game · 3 rounds · ~5 min

Exercise 1

Domain Detective

Guess the Domain

  • concepts vs instances
  • schemas
  • one vocabulary, three meanings
  • sets up T-Box vs A-Box

Facilitator-led slide · 3 rounds · keyboard driven

Exercise 2

Ontology Builder

Build an ontology, watch it render

  • C-Box terms before sorting
  • concepts, hierarchy, relations, attributes
  • instances, instance relations, values
  • the blueprint validates the data
  • exports W3C RDF / RDFS / SKOS / OWL as Turtle

Split-screen builder · palette of tools · guided 10 steps or free build

Exercise 3

Ask the Graph

SPARQL over the data and the blueprint

  • the triple as a question
  • one language for A-Box and T-Box
  • SKOS terms queried too
  • open standards, no vendor lock-in

Browser SPARQL workbench · ready-made questions · loads any Turtle

Business course

Facilitator-led exercises for the executive course. Run them full screen in a session.

Exercise 1

Domain Detective

Guess the Domain

  • concepts vs instances
  • schemas
  • one vocabulary, three meanings
  • sets up T-Box vs A-Box

Facilitator-led slide · 3 rounds · keyboard driven

Exercise 2

Ontology Builder

Build an ontology, watch it render

  • C-Box terms before sorting
  • concepts, hierarchy, relations, attributes
  • instances, instance relations, values
  • the blueprint validates the data
  • exports W3C RDF / RDFS / SKOS / OWL as Turtle

Split-screen builder · palette of tools · guided 10 steps or free build

Exercise 3

Ask the Graph

SPARQL over the data and the blueprint

  • the triple as a question
  • one language for A-Box and T-Box
  • SKOS terms queried too
  • open standards, no vendor lock-in

Browser SPARQL workbench · ready-made questions · loads any Turtle

Facilitator guides

Discussion prompts and notes to run after the checks and exercises.

Guide

Discussion Guide

After every check and exercise

  • prompts per exercise
  • what to listen for
  • closing three lines
  • prints with notes expanded

Facilitator page · pick 2–3 prompts per session