<div class="brand">eccenca<small>mastering complexity</small></div> # RDF and Linked Data Foundations eccenca Corporate Memory training · Module 1 Note: Welcome the participants. This module lays the foundation for everything that follows: SPARQL, ontology modeling and building Knowledge Graphs in Corporate Memory.

## Syntax and semantics <div class="columns"> <div> - **Syntax** – the *form* of the expressions of a language - **Semantics** – the *meaning* of the expressions of a language </div> <div> | Expression | Assessment | | --- | --- | | `4+)2=(` | incorrect syntax | | `4+2=10` | correct syntax, incorrect semantics | | `4+2=6` | correct syntax, correct semantics | | `4+2` vs. `2+4` | different syntax, same semantics | </div> </div> Note: Machines are good at syntax. The goal of the Semantic Web is to make meaning processable as well.

## Implicit and contextual meaning **Implicit meaning** – not stated, but derivable from something else > *Emil is a bear.* → *Emil is a mammal.* **Contextual meaning** – depends on the textual, situational or social surroundings > *The fisherman went to the bank.* – river bank or financial institution? Note: Ask the group how they knew that Emil is a mammal. That background knowledge is exactly what an ontology makes explicit.

## What a search engine does not find - *Apartments for rent close to well-rated Thai restaurants* - *Bilingual English-German childcare in Berlin, 15 minutes from my workplace* - *ERP service providers with offices in Vienna and Berlin* The knowledge is on the Web – but spread across sources, made for human reading, and without machine-readable meaning. **Needed:** structure, standards for representation and integration, and automated reasoning.

## One concept, many names <div class="columns"> <div> **System A:** "Plane"<br> **System B:** "Airplane"<br> **System C:** "Aeroplane" One shared concept – the *knowledge anchor*: generates lift · has wings · transports </div> <div> **Integrate knowledge** - Ontology: *Plane has part Wing* - Controlled vocabulary: *Plane = Airplane = Aeroplane* **Enrich knowledge** - Inference: *Airbus A320 has part Wing* - Explainable input for AI and LLMs </div> </div>

## The Semantic Web **Basic ingredients** - Open standards for representing information and meaning on the Web - Methods for deriving further information from such descriptions **The role of ontologies** - A reference frame for disambiguation and global linking of meaning - Background and contextual knowledge – implicit knowledge becomes derivable - Formal semantics with an inference framework, shared via the Web

The Semantic Web stack

User interface and applications

Trust

Proof

Crypto

Unifying logic

Query: SPARQL

Ontology: OWL

Rules: RIF

RDF Schema

Data interchange: RDF

XML

URI / IRI

Unicode

## Linked Data principles Tim Berners-Lee, 2006: 1. Use **URIs** as names for things. 2. Use **HTTP URIs** so that people can look up those names. 3. When someone looks up a URI, provide **useful information**, using standards such as RDF and SPARQL. 4. Include **links to other URIs**, so that they can discover more things. <p class="small muted">The Linked Open Data cloud grew from a dozen datasets in 2007 to more than a thousand within ten years – DBpedia, GeoNames, Wikidata, life science and government data.</p>

<div class="brand">eccenca<small>mastering complexity</small></div> # The RDF data model

## Resource Description Framework (RDF) - A **data model** – the basic representation format of the Semantic Web - W3C Recommendation since 1999, current version **RDF 1.1** (2014) - First used for metadata about web resources, then generalized - A universal, machine-readable exchange format for structured information - Queried with **SPARQL** – see the next module

Parts of an RDF graph

• Data is a graph of nodes and directed edges
• IRIs name nodes and edges unambiguously
• Literals are data values: strings, numbers, dates
• Blank nodes stand for a resource without a name

An RDF graph

## RDF statements: triples - RDF consists of **statements** - A statement is a **triple**: subject – predicate – object | Position | Allowed | | --- | --- | | Subject | IRI or blank node | | Predicate | IRI | | Object | IRI, blank node or literal | > **Burkhard Jung** – *isMayorOf* – **Leipzig** A graph is the set of its triples – it can be reconstructed from the triple list.

## IRIs – Internationalized Resource Identifiers - Globally unique names for resources – anything with a clear identity - Like an ISBN for books, but for everything - Structure: `scheme:[//[user[:password]@]host[:port]][/path][?query][#fragment]` - A name first – it need not resolve to a document - Own IRIs belong in an HTTP namespace under own control ```text http://dbpedia.org/resource/Leipzig http://ld.company.org/prod-vocab/Product ```

## Literals - Model **data values**, written as strings - Interpreted through a **datatype**, usually from XML Schema: `xsd:decimal`, `xsd:date`, … - Without a datatype, a literal is an `xsd:string` - Strings can carry a **language tag**: `@en`, `@de`, `@pt` - Literals only appear in the **object** position – never as subject, never as edge

## Typed literals Untyped values sort as text: `"02" < "100" < "11" < "2"` ```turtle @prefix dbr: <http://dbpedia.org/resource/> . @prefix geo: <http://www.w3.org/2003/01/geo/wgs84_pos#> . @prefix xsd: <http://www.w3.org/2001/XMLSchema#> . dbr:Leipzig geo:long "12.3833"^^xsd:decimal ; geo:lat "51.3333"^^xsd:decimal . ``` Syntax: `"value"^^<datatype-IRI>` – `"Leipzig"` is short for `"Leipzig"^^xsd:string`

## Labels and language tags `rdfs:label` gives a resource a human-readable name. ```turtle @prefix dbr: <http://dbpedia.org/resource/> . @prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> . dbr:Leipzig rdfs:label "Leipzig"@de , "Lipsia"@it , "Лейпциг"@ru . ``` <p class="small muted">Language tags follow BCP 47 (RFC 5646).</p>

## Serializations One graph, many syntaxes: | Format | Character | | --- | --- | | **N-Triples** | one triple per line, full IRIs – easy to process | | **Turtle** | compact and human-readable – used in this training | | **JSON-LD** | RDF as JSON – popular with web developers | | **RDF/XML** | the original XML syntax | | **TriG**, **N-Quads** | several named graphs in one file |

## Turtle: the basics - IRIs in angle brackets: `<http://dbpedia.org/resource/Leipzig>` - Literals in quotes: `"Leipzig"`, `"51.3333"` - A triple is a sentence: subject, predicate, object, **terminated with a dot** - Whitespace and line breaks outside of terms do not matter ```turtle <http://dbpedia.org/resource/Leipzig> <http://dbpedia.org/property/locatedIn> <http://dbpedia.org/resource/Saxony> . ```

## Turtle: prefixes `@prefix abbr: <namespace-IRI> .` shortens IRIs to `abbr:localName` ```turtle <http://dbpedia.org/resource/Leipzig> <http://dbpedia.org/property/locatedIn> <http://dbpedia.org/resource/Saxony> . ``` is the same as ```turtle @prefix dbr: <http://dbpedia.org/resource/> . @prefix dbp: <http://dbpedia.org/property/> . dbr:Leipzig dbp:locatedIn dbr:Saxony . ```

## Turtle: `;` and `,` - `;` – next predicate for the **same subject** - `,` – next object for the **same subject and predicate** ```turtle @prefix dbr: <http://dbpedia.org/resource/> . @prefix dbp: <http://dbpedia.org/property/> . @prefix xsd: <http://www.w3.org/2001/XMLSchema#> . dbr:Leipzig dbp:hasAreaCode "0341" ; dbp:locatedIn dbr:Saxony , dbr:Germany ; dbp:hasMayor dbr:Burkhard_Jung . dbr:Burkhard_Jung dbp:born "1958-03-07"^^xsd:date ; dbp:isMemberOf dbr:Social_Democratic_Party_of_Germany . ```

## Classes - Resources belong to **classes** – a class is a set of things, named by an IRI - `rdf:type` states membership – Turtle shorthand: `a` ```turtle ex:MyBlueVWGolf rdf:type ex:Car . ex:MyBlueVWGolf a ex:Car . # the same triple ``` - A resource can belong to several classes ```turtle ex:MyBlueVWGolf a ex:Car , ex:GermanProduct . ``` <p class="small muted">Class hierarchies, domains and ranges follow in the ontology module.</p>

## n-ary relations: the problem <span class="ask">For mango chutney, take 450 g of green mango and a teaspoon of cayenne pepper …</span> ```turtle ex:Chutney ex:hasIngredient "450g green mango" , "1tsp cayenne pepper" . ``` → ingredients hidden in strings, no search for "recipes with green mango" ```turtle ex:Chutney ex:ingredient ex:GreenMango ; ex:amount "450g" ; ex:ingredient ex:CayennePepper ; ex:amount "1tsp" . ``` → which amount belongs to which ingredient?

## n-ary relations: helper nodes <div class="columns"> <div> | Recipe | Ingredient | Amount | | --- | --- | --- | | Mango chutney | Green mango | 450 g | | Mango chutney | Cayenne pepper | 1 tsp | A three-way relation – RDF only has two-way edges. </div> <div> ```turtle ex:Chutney ex:hasIngredient ex:ChutneyIngredient1 . ex:ChutneyIngredient1 ex:ingredient ex:GreenMango ; ex:amount "450g" . ``` </div> </div>

## Blank nodes - For resources that need no global name – read as "there is something that …" - Label `_:id1` is only valid inside one document ```turtle ex:Chutney ex:hasIngredient _:id1 . _:id1 ex:ingredient ex:GreenMango ; ex:amount "450g" . ``` Shorter, with square brackets: ```turtle ex:Chutney ex:hasIngredient [ ex:ingredient ex:GreenMango ; ex:amount "450g" ] . ```

## Lists: containers and collections <div class="columns"> <div> **Containers** – open - `rdf:Bag` – unordered - `rdf:Seq` – ordered - `rdf:Alt` – alternatives ```turtle ex:AKSW ex:members [ a rdf:Bag ; rdf:_1 ex:Sören ; rdf:_2 ex:Jens ] . ``` </div> <div> **Collections** – closed, ordered ```turtle ex:AKSW ex:groupLeaders ( ex:Sören ex:Jens ex:Axel ) . ``` Built from `rdf:first`, `rdf:rest` and `rdf:nil` – a linked list. </div> </div>

## Summary - RDF is a widely supported standard for storing and exchanging data - Triple stores handle very large RDF graphs - Information from many sources merges into one graph – a **Knowledge Graph** **Next** - How to describe *schema* knowledge – classes, properties, hierarchies → **Ontology modeling** - How to *retrieve* information from a graph → **SPARQL basics**