3. Everything is a triple | Semantic Layer course

Section 1

One shape for every fact

Next section

Section 2

Why identifiers look like web addresses

Next section

Section 3

Values, types, and Turtle

Next section

Section 4

One row becomes eleven triples

Next section

Section 5

Adding a fact nobody planned for

Next section

Section 6

Named graphs: which triples do you trust?

Hands on

Load Priya's mortgage into a triple store

20 min

Write valid Turtle for customer C-1001 and loan L-1042, load it into Apache Jena Fuseki, and verify the triple count with a query.

make lab-up            # once, starts Postgres + Fuseki + Ontop
make lab lab03-first-triples

Full instructions, starter files and the solution are in labs/modules/lab03-first-triples/README.md. The lab checks its own work, so you will know if it passed.

Check yourself

Quiz

6 questions, pass mark 70%. Answers are explained either way, so a wrong guess is not wasted.

1 Why does RDF identify Priya's mortgage as https://meridianbank.example/data/Loan/L-1042 rather than just L-1042?

Because triple stores require the identifier to be fetchable over HTTP Because the IRI is globally unique, so two systems using it are provably talking about the same loan Because longer identifiers compress better Because SPARQL cannot query short strings

2 In Turtle, what is the difference between a semicolon and a comma at the end of a line?

A semicolon repeats the subject; a comma repeats both the subject and the predicate A semicolon ends a triple; a comma starts a comment They are interchangeable A comma repeats the subject; a semicolon repeats the predicate

3 Which two statements about literals are true? (pick two)

Select all that apply.

A literal can be the subject of another triple, so you can state facts about a value Money should be typed xsd:decimal rather than a floating-point type A bare quoted value with no datatype marker is treated as xsd:string Language tags and datatypes can both be attached to the same literal

4 Meridian needs to record that a risk rating was assigned by model version v3 on 2026-02-01. What does this cost in each model?

An ALTER TABLE plus a 4.1 million row backfill in SQL, versus writing two more triples in RDF A new table in both models Nothing in SQL, since it can go in a JSON column, and a new class in RDF A schema migration in both models

5 Loan L-4590 has mrb:daysPastDue 104 loaded from loan_master, and a reasoner then derives that it is a NonPerformingLoan. Why put those two triples in different named graphs?

Because inferred triples use a different syntax So you can report which facts were stated versus derived, and rebuild all derived facts by dropping one graph when a rule changes Because Fuseki cannot store them together To make queries faster by reducing the number of triples scanned

6 Why does this course tell you to avoid blank nodes at Meridian?

Turtle does not support them They are slower to query in every triple store You cannot reference them from another file, cite them in an audit trail, or reliably update them, because their label can change between loads They are only valid inside named graphs