Distributed Knowledge Graphs 1

UNC SDIS, INLS 620, Fall 2026

August 18
Introduction to the course

Total amount of required reading for this meeting: 1,000 words

An introduction to the course: what knowledge graphs are, what makes them distributed, and how we will work together this semester.

๐Ÿ“– To read before this meeting:

  1. Recurse Center. “Social Rules,” n.d. https://www.recurse.com/social-rules.
    1,000 words

August 20
Git and GitHub

View slides Updated Thursday 8/20 10:36 AM

Git is a distributed version control system, and GitHub hosts Git repositories on the Web. Each assignment is distributed as a repository that is forked for you; we look at how to clone your fork, and how to stage, commit, and push your work back to it.

๐Ÿ“– To read before this meeting:

  1. GitHub, Inc. “Start Your Journey: Learn the Basics of GitHub.” GitHub Docs, 2024. https://docs.github.com/en/get-started/start-your-journey.

August 24
Hello world assignment available

August 25
Visual Studio Code and Codespaces

View slides Updated Tuesday 8/25 10:31 AM

Total amount of required viewing for this meeting: 14 minutes

Visual Studio Code is the editor we will use throughout the course. A codespace is a preconfigured Linux development environment that runs in a container, either remotely on GitHub or locally using Docker.

๐Ÿ“บ To view before this meeting:

  1. Microsoft, โ€œLearn Visual Studio Code in 7min.โ€

    View. 7 minutes
    Viewing tips

    You can follow along at https://vscode.dev.

  2. Microsoft, โ€œUsing Git with Visual Studio Code.โ€

    View. 7 minutes

August 27
Forms of data

View slides Updated Wednesday 8/26 8:13 PM

Total amount of required reading for this meeting: 12,800 words

Data can be structured as blobs, sets, lists, dictionaries, trees, or graphs, and each of these structures makes different questions easy to ask. We compare the tabular, relational, meta-markup, and graph metamodels built on top of them.

๐Ÿ“– To read before this meeting:

  1. Shaw, Ryan. “The Forms of Descriptions (Part 1),” 2022. PDF.
    4,300 words
  2. Shaw, Ryan. “The Forms of Descriptions (Part 2),” 2022. PDF.
    8,500 words

August 27
Hello world assignment due

September 1
(Semantic) data modeling I

View slides Updated Tuesday 9/1 4:16 PM

Total amount of required viewing for this meeting: 19 minutes

Data modeling is the design of a formal language intended to aid communication and mediate among different purposes and perspectives.

๐Ÿ“บ To view before this meeting:

  1. Posner, Miriam. What Is Linked Open Data?, 2021.

    View. 19 minutes

September 3
(Semantic) data modeling II

View slides Updated Thursday 9/3 1:49 PM

Total amount of required reading for this meeting: 8,000 words

The Semantic Web is a vision of the Web as a space for sharing data rather than documents. We consider what it means to model data so that it can be combined with data modeled by other people for other purposes.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “What Is the Semantic Web?” In Semantic Web for the Working Ontologist, 3rd ed., 1–18. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    8,000 words

September 3
Data modeling assignment available

September 8
(Semantic) data modeling III

View slides Updated Tuesday 9/8 1:57 PM

Total amount of required reading for this meeting: 7,000 words

Semantic modeling makes the meaning of data explicit, so that data can be interpreted consistently by parties who never coordinated with one another.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “Semantic Modeling.” In Semantic Web for the Working Ontologist, 3rd ed., 19–35. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    7,000 words

September 10
The Web I

View slides Updated Thursday 9/10 1:55 PM

Total amount of required reading for this meeting: 10,200 words

The (World Wide) Web is an Internet-scale distributed hypermedia system. It exists by means of voluntary compliance with open communication protocols and data format standards.

๐Ÿ“– To read before this meeting:

  1. Richardson, Leonard, and Mike Amundsen. “Surfing the Web.” In RESTful Web APIs, 1–16. Sebastopol, California: O’Reilly, 2013. PDF.
    4,500 words
  2. Hogan, Aidan. “Introduction.” In The Web of Data, 1–14. Springer, 2020. PDF.
    5,700 words
    Reading tips

    Focus on section 1.2 (โ€œThe Current Webโ€), pages 6โ€“14.

September 15
The Web II

View slides Updated Tuesday 9/15 1:52 PM

Total amount of required reading for this meeting: 4,900 words

A closer look at the machinery of the Web: resources identified by URIs, representations transferred over HTTP, and the roles played by DNS, HTTP methods, status codes, and content negotiation.

๐Ÿ“– To read before this meeting:

  1. Richardson, Leonard, and Mike Amundsen. “Resources and Representations.” In RESTful Web APIs, 29–43. Sebastopol, California: O’Reilly, 2013. PDF.
    4,900 words
  2. Optional
    Allemang, Dean, Jim Hendler, and Fabien Gandon. “Linked Data.” In Semantic Web for the Working Ontologist, 3rd ed., 85–118. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    11,600 words

September 15
Data modeling assignment due

September 15
The Web assignment available

September 17
RDF: terms, triples, graphs I

View slides Updated Thursday 9/17 5:25 PM

Total amount of required reading for this meeting: 7,000 words

The Resource Description Framework (RDF) is a conceptual model for structuring information into triples that can be combined into graphs.

๐Ÿ“– To read before this meeting:

  1. DuCharme, Bob. “What Is RDF?,” June 27, 2021. https://www.bobdc.com/blog/whatisrdf/.
    2,100 words
  2. Schreiber, Guus, and Yves Raimond. “RDF Data Model.” In RDF 1.1 Primer. W3C, 2014. https://www.w3.org/TR/rdf11-primer/#section-data-model.
    1,300 words

September 22
RDF: terms, triples, graphs II

View slides Updated Tuesday 9/22 2:53 PM

Total amount of required reading for this meeting: 7,000 words

The three kinds of node found in an RDF graphโ€”URIs, literals, and blank nodesโ€”and the properties that relate them. We also look at namespaces and CURIEs, and at how RDF datasets keep track of graphs that change over time.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “RDF—The Basis of the Semantic Web.” In Semantic Web for the Working Ontologist, 3rd ed., 37–67. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    7,000 words
  2. Optional
    Hogan, Aidan. “Resource Description Framework.” In The Web of Data, 59–109. Springer, 2020. PDF.
    10,000 words
  3. Optional
    W3C. “RDF 1.1 Concepts and Abstract Syntax,” February 25, 2014. https://www.w3.org/TR/rdf11-concepts/.

September 24
RDF: serializations I

View slides Updated Thursday 9/24 3:00 PM

Total amount of required reading for this meeting: 5,100 words

Because RDF is a purely conceptual model, it does not specify how data should be written down or serialized. There are several alternative standards for serialization.

๐Ÿ“– To read before this meeting:

  1. Schreiber, Guus, and Yves Raimond. “Writing RDF Graphs.” In RDF 1.1 Primer. W3C, 2014. https://www.w3.org/TR/rdf11-primer/#section-graph-syntax.
    2,900 words
  2. Allemang, Dean, Jim Hendler, and Fabien Gandon. “Alternatives for Serialization.” In Semantic Web for the Working Ontologist, 3rd ed., 68–63. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    2,200 words

September 29
RDF: serializations II

View slides Updated Tuesday 9/29 4:31 PM

Total amount of required reading for this meeting: 5,000 words

A survey of the standard RDF syntaxesโ€”N-Triples, Turtle, RDF/XML, and JSON-LDโ€”and the tradeoffs among them, including what JSON-LD offers to developers already working with JSON.

๐Ÿ“– To read before this meeting:

  1. Hogan, Aidan. “RDF Syntaxes.” In The Web of Data, 94–109. Springer, 2020. PDF.
    5,000 words
  2. Optional
    Sporny, Manu. What Is JSON-LD?, 2012. https://www.youtube.com/watch?v=vioCbTo3C-4.

September 29
The Web assignment due

October 1
RDF Schema and RDF vocabularies I

View slides Updated Thursday 10/1 2:54 PM

Total amount of required reading for this meeting: 2,200 words

An RDF vocabulary is a set of IRIs and the triples that describe them, and RDF Schema is the vocabulary for writing those triples. We declare classes and properties, document them with labels and comments, look terms up by asking their IRIs, and weigh reusing a published vocabulary against minting our own.

๐Ÿ“– To read before this meeting:

  1. DuCharme, Bob. “What Is RDFS?,” July 25, 2021. https://www.bobdc.com/blog/whatisrdfs/.
    1,500 words
  2. Schreiber, Guus, and Yves Raimond. “RDF Vocabularies.” In RDF 1.1 Primer. W3C, 2014. https://www.w3.org/TR/rdf11-primer/#section-vocabulary.
    700 words

October 6
Well-being day

Due to the well-being day, we will not meet today.

October 8
RDF Schema and RDF vocabularies II

View slides Updated Thursday 10/8 3:34 PM

Total amount of required reading for this meeting: 11,300 words

RDF Schema gives a vocabulary meaning through inference: the subclass and subproperty relations and the domains and ranges of properties let new triples be derived from asserted ones. We see why RDFS never rejects data, and how the same rules map one vocabulary onto another.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “RDF Schema.” In Semantic Web for the Working Ontologist, 3rd ed., 201–32. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    11,300 words
  2. Optional
    Hogan, Aidan. “RDF Schema and Semantics.” In The Web of Data, 111–83. Springer, 2020. PDF.
    25,100 words
  3. Optional
    W3C. “RDF Schema 1.1,” February 25, 2014. https://www.w3.org/TR/rdf-schema/.
    4,000 words

October 8
RDFS assignment available

October 13
Wikidata and Wikibase I

View slides Updated Tuesday 10/13 2:53 PM

Total amount of required reading for this meeting: 5,100 words

An introduction to Wikidata through its web pages: items and properties, the statements that link them, and the references that back them up. Wikidata has its own versions of rdf:type, rdfs:subClassOf, and domain and range, but where RDFS infers, Wikidata only reports.

๐Ÿ“– To read before this meeting:

  1. Vanderbilt University. “Learn Wikidata.” Accessed January 7, 2022. https://www.learnwikidata.net/.
  2. Vrandečić, Denny, and Markus Krötzsch. “Wikidata: A Free Collaborative Knowledgebase.” Communications of the ACM 57, no. 10 (September 23, 2014): 78–85. https://doi.org/10.1145/2629489.
    5,100 words

October 15
Fall break

Due to fall break, we will not meet today.

October 20
Wikidata and Wikibase II

View slides Updated Tuesday 10/20 3:45 PM

Total amount of required reading for this meeting: 6,800 words

Wikidata as RDF: each statement is a node, with its qualifiers, references and rank attached, and the short wdt: triples are derived from the best-ranked statements. Values come in a simple form and a full one, and unknown value and no value are each said explicitly.

๐Ÿ“– To read before this meeting:

  1. Wikidata. “Help:Items,” n.d. https://www.wikidata.org/wiki/Help:Items.
    1,900 words
  2. Wikidata. “Help:Properties,” n.d. https://www.wikidata.org/wiki/Help:Properties.
    1,200 words
  3. Wikidata. “Help:Statements,” n.d. https://www.wikidata.org/wiki/Help:Statements.
    1,800 words
  4. Wikidata. “Help:Data Type,” n.d. https://www.wikidata.org/wiki/Help:Data_type.
    1,900 words
  5. Optional
    Wikibase. “Wikibase/DataModel,” n.d. https://www.mediawiki.org/wiki/Wikibase/DataModel.
    7,200 words
  6. Optional
    Wikibase. “Wikibase/Indexing/RDF Dump Format,” n.d. https://www.mediawiki.org/wiki/Wikibase/Indexing/RDF_Dump_Format.
    4,800 words

October 20
RDFS assignment due

October 22
SPARQL I

View slides Updated Thursday 10/22 3:23 PM

Total amount of required reading for this meeting: 19,200 words

SPARQL queries RDF by matching graph patterns, triples with variables in place of some of their terms. This session covers joining patterns through a shared variable, narrowing results with FILTER, DISTINCT, ORDER BY and LIMIT, and handling alternatives and missing data with UNION, OPTIONAL and FILTER NOT EXISTS.

๐Ÿ“– To read before this meeting:

  1. DuCharme, Bob. “Jumping Right in: Some Data and Some Queries.” In Learning SPARQL: Querying and Updating with SPARQL 1.1, 2nd ed., 1–17. Sebastopol: O’Reilly Media, 2013. PDF.
    5,100 words
  2. DuCharme, Bob. “SPARQL Queries: A Deeper Dive.” In Learning SPARQL: Querying and Updating with SPARQL 1.1, 2nd ed., 47–102. Sebastopol: O’Reilly Media, 2013. PDF.
    14,100 words
  3. Shaw, Ryan. “Learning SPARQL, Chapters 1 and 3: The Queries, Linked,” 2026. PDF.

October 27
Ryan is out of town

I will be be out of town, so we will not meet today.

October 29
SPARQL II

View slides Updated Thursday 10/29 2:56 PM

Total amount of required viewing for this meeting: 45 minutes

Total amount of required reading for this meeting: 2,200 words

Querying Wikidata with SPARQL: the label service, the wdt:, p: and ps: predicates from Wikidata and Wikibase II, and property paths such as wdt:P31/wdt:P279*, which follow a chain of triples of any length.

๐Ÿ“บ To view before this meeting:

  1. Wikimedia Foundation. Querying Wikidata with SPARQL for Absolute Beginners, 2018.

    View. 45 minutes
    Viewing tips

    You need not watch the entire 111 minutesโ€ฆ the first 45 minutes covers the most important things.

๐Ÿ“– To read before this meeting:

  1. “A Gentle Introduction to the Wikidata Query Service.” In Wikidata, n.d. https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/A_gentle_introduction_to_the_Wikidata_Query_Service.
    2,200 words
  2. Optional
    Wikidata. “Wikidata:SPARQL Query Service/Queries,” n.d. https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/queries.

October 29
Wikidata and SPARQL assignment available

November 3
Ryan was ill

I was ill, so we did not meet today.

November 5
SPARQL III

View slides Updated Thursday 11/5 3:08 PM

Total amount of required reading for this meeting: 18,900 words

Grouping results and computing aggregates such as counts and averages for each group, filtering groups with HAVING, and CONSTRUCT queries that return graphs rather than tables. Part of the session is set aside for questions about the SPARQL assignment.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “Querying the Semantic Web—SPARQL.” In Semantic Web for the Working Ontologist, 3rd ed., 119–80. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    18,900 words
  2. Optional
    Hogan, Aidan. “SPARQL Query Language.” In The Web of Data, 323–448. Springer, 2020. PDF.
    37,500 words
  3. Optional
    W3C. “SPARQL 1.1 Query Language,” March 21, 2013. https://www.w3.org/TR/sparql11-query/.

November 10
Shape constraints I

View slides Updated Tuesday 11/10 3:17 PM

Total amount of required reading for this meeting: 7,400 words

A shape describes what we expect a graph to contain, and validation checks a graph against shapes, reporting what it lacks rather than inferring what it might mean, the opposite of what RDF Schema does. The Shapes Constraint Language (SHACL) writes shapes in RDF: node shapes and property shapes, the targets that pick out which nodes to check, and the validation report that says which nodes fail and why.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “Extending RDF: RDFS and SHACL.” In Semantic Web for the Working Ontologist, 3rd ed., 181–200. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    7,400 words

November 10
Wikidata and SPARQL assignment due

November 12
Shape constraints II

View slides Updated Thursday 11/12 3:46 PM

Total amount of required reading for this meeting: 19,800 words

SHACL is grounded in SPARQL: its paths are SPARQL property paths, each of its constraints amounts to a query for the nodes that violate it, and SHACL-SPARQL lets us write such queries ourselves when the built-in constraints aren’t enough. We also combine and close shapes, and look briefly at ShEx, the alternative shapes language that Wikidata uses for its entity schemas.

๐Ÿ“– To read before this meeting:

  1. Hogan, Aidan. “Shape Constraints and Expressions.” In The Web of Data, 449–513. Springer, 2020. PDF.
    19,800 words
    Reading tips

    Focus on section 7.1 on SHACL (pages 453โ€“501), skimming the catalogue of core constraints in 7.1.4 and reading 7.1.5 on SHACL-SPARQL (pages 492โ€“500) closely, and on section 7.3 comparing SHACL and ShEx (page 508). Section 7.2 on ShEx is optional.

  2. Optional
    Wikidata. “Wikidata:WikiProject Schemas,” n.d. https://www.wikidata.org/wiki/Wikidata:WikiProject_Schemas.

November 12
SHACL and OWL assignment available

November 17
Shape constraints III / OWL I

View slides Updated Tuesday 11/17 2:47 PM

Total amount of required reading for this meeting: 25,000 words

Many jobs the Web Ontology Language (OWL) was first used forโ€”documenting what data should look like, checking it, generating forms for itโ€”are now increasingly done with SHACL, leaving OWL the job of drawing inferences from incomplete data. We then begin OWL with RDFS-Plus: inverse, transitive, and symmetric properties and the distinction between object and datatype properties, as used to define the Simple Knowledge Organization System (SKOS) and as used in the wild.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “RDFS-Plus.” In Semantic Web for the Working Ontologist, 3rd ed., 233–69. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    12,400 words
  2. Allemang, Dean, Jim Hendler, and Fabien Gandon. “Using RDFS-Plus in the Wild.” In Semantic Web for the Working Ontologist, 3rd ed., 271–302. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    12,600 words
    Reading tips

    Focus on sections 10.1 on Schema.org (pages 273โ€“280) and 10.3 on FOAF (pages 289โ€“298).

  3. W3C Data Shapes Working Group. “SHACL 1.2 Overview.” W3C. Accessed November 8, 2026. https://w3c.github.io/data-shapes/shacl12-overview/.
    Reading tips

    An Editor’s Draft, so it may change; read โ€œWhat’s New in SHACL 1.2โ€ and โ€œSHACL-OWLโ€ for how SHACL is taking on rules, classes, and user interfaces.

  4. Optional
    Allemang, Dean, Jim Hendler, and Fabien Gandon. “SKOS—Managing Vocabularies with RDFS-Plus.” In Semantic Web for the Working Ontologist, 3rd ed., 303–18. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    5,700 words

November 19
OWL II

View slides Updated Thursday 11/19 5:25 PM

Total amount of required reading for this meeting: 10,000 words

OWL lets us say when two names refer to the same thing (owl:sameAs, functional and inverse-functional properties), and define classes by restrictions on the values their members have for a property (owl:hasValue, owl:someValuesFrom, owl:allValuesFrom), so that a reasoner can classify individuals from what is known about them.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “Basic OWL.” In Semantic Web for the Working Ontologist, 3rd ed., 319–51. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    10,000 words

November 24
OWL III

Total amount of required reading for this meeting: 11,500 words

Defining classes by set operations (union, intersection, and complement) and cardinality restrictions, declaring classes disjoint and individuals distinct, and what a reasoner does with a contradiction. Setting an OWL cardinality restriction beside the SHACL constraint that looks the same shows the difference between inferring from incomplete data and validating it.

๐Ÿ“– To read before this meeting:

  1. Allemang, Dean, Jim Hendler, and Fabien Gandon. “Counting and Sets in OWL.” In Semantic Web for the Working Ontologist, 3rd ed., 353–89. ACM Books 33. Association for Computing Machinery, 2020. PDF.
    11,500 words
  2. Optional
    Hogan, Aidan. “Web Ontology Language.” In The Web of Data, 185–322. Springer, 2020. PDF.
    49,300 words

November 26
Thanksgiving

Due to Thanksgiving, we will not meet today.

December 1
Semantic data modeling in practice

Guest speaker Derek Rodriguez, Ph.D., Managing Director and Practice Lead for Global Knowledge Engineering at Accenture AI & Data, on semantic data modeling as it is done in industry. Derek holds an MSLS and a PhD from UNC SILS (now SDIS).