8. The virtual knowledge graph | Semantic Layer course

Section 1

The awkward question after Module 7

Next section

Section 2

Two ways to have a graph

Next section

Section 3

R2RML, the mapping language

Next section

Section 4

When the table does not cooperate

Next section

Section 5

SPARQL in, SQL out

Next section

Section 6

More than one source

Next section

Section 7

Who is allowed to see which rows

Hands on

Query Postgres as a graph

40 min

Load Meridian's loan tables into Postgres, write the R2RML mapping, start Ontop as a SPARQL endpoint over it, then run the same SPARQL query from Module 4 and get the same answer with no data having moved.

make lab-up            # once, starts Postgres + Fuseki + Ontop
make lab lab08-ontop-vkg

Full instructions, starter files and the solution are in labs/modules/lab08-ontop-vkg/README.md. The lab checks its own work, so you will know if it passed.

Check yourself

Quiz

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

1 Which two of Meridian's data assets should be materialized rather than left virtual? (pick two)

Select all that apply.

The 52,400 rows in loan_master, which the core banking system updates nightly The roughly 800 facts extracted from loan documents, which came from no table The triples the OWL reasoner inferred, including the non-performing classifications The 2 million rows in the daily payment transaction table

2 In an R2RML mapping, what does rr:template on a subject map produce?

The SQL query that selects the rows The IRI for each row's subject, built by substituting column values into a URL pattern A SHACL shape that validates the row The datatype of the object literal

3 You need loan:L-1042 mrb:borrower cust:C-1001 built from the cust_ref foreign key. Which R2RML construct does it?

An rr:objectMap with rr:column "cust_ref" An rr:objectMap with rr:parentTriplesMap and rr:joinCondition naming rr:child and rr:parent An rr:graphMap with rr:constant A second rr:subjectMap on the same triples map

4 A SPARQL query has four triple patterns on the same ?loan variable, all mapping to loan_master. Ontop emits SQL with one reference to loan_master. Which optimization did that?

Predicate pushdown Filter propagation Self-join elimination, because loan_id is unique so the joins cannot change the result Join elimination on the customer foreign key

5 Why does the guidance warn that auto-generated R2RML mappings are only a draft?

Generators produce invalid Turtle that will not parse Column names carry meaning unevenly, so a generator turns stat_cd into a string property with no idea that code 03 means non-performing R2RML cannot express foreign keys, so those must always be hand written Generated mappings only work against MySQL

6 Which enforcement strategy avoids keeping a second copy of the access policy in sync with the first?

SPARQL query rewriting that injects FILTER clauses Trusted identity propagation through IAM Identity Center, so Lake Formation enforces natively at the source SQL predicate injection into the generated query Restricting the SPARQL endpoint to an internal subnet

7 A JWT reaching the endpoint has sub set to a human analyst and act set to a loan Q&A agent. Why does the act claim matter?

It selects which row filters apply, replacing the sub claim It records that an agent made the call on the human's behalf, so 4,000 overnight queries read as automation rather than a stolen credential It carries the SPARQL query text for auditing It grants the agent broader access than the human has