What is the difference between RDF and OWL?

A picture speaks a thousand words! This diagram below should reinforce what Christopher Gutteridge said in this answer that the semantic web is a "layered architecture".

Source: https://www.obitko.com/tutorials/ontologies-semantic-web/semantic-web-architecture.html

The basic semantic web stack has been explained a lot already in this thread. I'd like to focus on the initial question and compare RDF to OWL.

• OWL is a super-set of RDF & RDF-S (on top)
• OWL allows to effectively working with RDF & RDF-S
• OWL has some extended vocabulary
  • classes & individuals ("instances")
  • properties & data-types ("predicates")
• OWL is required for proper reasoning and inference
• OWL comes in three dialects lite, description logic & full

Using OWL is essential to get more meaning (reasoning & inference) by just knowing a few facts. This "dynamically created" information can further be used for accordant queries like in SPARQL.

Some examples will show that that actually works with OWL - these have been taken from my talk about the basics of semantic web at the TYPO3camp Mallorca, Spain in 2015.

equivalents by rules

Spaniard: Person and (inhabitantOf some SpanishCity)

This means that a Spaniard must be a Person (and thus inherits all properties in the inferencing part) and must live in at least one (or more) SpanishCity.

meaning of properties

<Palma isPartOf Mallorca>
<Mallorca contains Palma>

The example shows the result of applying inverseOf to the properties isPartOf and contains.

• inverse
• symmetric
• transitive
• disjoint
• ...

cardinalities of properties

<:hasParent owl:cardinality “2“^^xsd:integer>

This defines that each Thing (in this scenario most probably a Human) has exactly two parents - the cardinality is assigned to the hasParent property.

• minimum
• maximum
• exact

The Resource Description Framework (RDF) is a powerful formal knowledge representation language and a fundamental standard of the Semantic Web. It has its own vocabulary that defines core concepts and relations (e.g., rdf:type corresponds to the isA relationship), and a data model that enables machine-interpretable statements in the form of subject-predicate-object (resource-property-value) triples, called RDF triples, such as picture-depicts-book. The extension of the RDF vocabulary with concepts required to create controlled vocabularies and basic ontologies is called RDF Schema or RDF Vocabulary Description Language (RDFS). RDFS makes it possible to write statements about classes and resources, and express taxonomical structures, such as via superclass-subclass relationships.

Complex knowledge domains require more capabilities than what is available in RDFS, which led to the introduction of OWL. OWL supports relationships between classes (union, intersection, disjointness, equivalence), property cardinality constraints (minimum, maximum, exact number, e.g., every person has exactly one father), rich typing of properties, characteristics of properties and special properties (transitive, symmetric, functional, inverse functional, e.g., A ex:hasAncestor B and B ex:hasAncestor C implies that A ex:hasAncestor C), specifying that a given property is a unique key for instances of a particular class, and domain and range restrictions for properties.