Semantic Technologies

Semantic technologies are a class of data management technologies that focus on making the underlying meaning of data more transparent to both human and machine users.

Semantic technologies encompass a variety of standards, frameworks, data models, and applications, which are described in further detail below.

Semantic Data Models

Semantic technologies organize information using the following data models.

Taxonomies

A taxonomy consists of hierarchies of concepts. For example, the concept poodle is a subclass of the concept dog, in the sense that all poodles are dogs. In turn, the concept dog is a subclass of the concept mammal. Taxonomies are data models consisting of definitions of concepts and a specification of the hierarchies (relations of subclass and superclass) between them.

Ontologies

An ontology is an enriched taxonomy. Ontologies include all the elements in taxonomies and add the following:

• Additional relations over concepts: Ontologies can include relations other than subclass and superclass. For example, to represent the information that cars have tires as parts, one can define a relation partOf as holding between the concepts tire and car. Notice that this is not a hierarchy relation: it does not say that all tires are cars or that all cars are tires.
• Properties of concepts: Ontologies can include properties that apply to concepts. For example, to represent the information that cars have weights, one can define the property weight as applying to the concept car.
• Additional information about properties and relations
  • The domain and range of properties and relations: Ontologies can include specifications of what types of entities a property or relation applies to (domain) and what values it can take (range). For example, one could specify that the domain of a relation authorOf is person, and the range is document.
  • Formal structures of properties and relations: Ontologies can include axioms specifying the formal structures of properties and relations. For example, one could include an axiom that partOf is transitive, to convey the information that if A is partOf B, and B is partOf C, then A is partOf C.

Knowledge Graphs

A knowledge graph models the properties of, and relations between, particular things using an ontology. For example, suppose a grocery store had an ontology modeling all of the concepts relevant to their operations, including concepts about store employees (cashier, manager, employee, etc.), products (napkins, produce, apple), and so on. They could then create a knowledge graph which models and tracks the relationships between particular employees (Bob Johnson, Snoop Dog), products as identified by barcodes (175937462846, 175937462840), and so on.

Fundamental Semantic Technology Stack

Most semantic technologies are based on the three following fundamental standards.

Resource Description Framework (RDF)

RDF is a language for storing data. RDF stores data in triples, which represent information using a subject-predicate-object format. For example, the information that Brazil is larger than Italy could be represented as a triple as follows:

subject:Brazil predicate:largerThan object:Italy

Note: This example is not valid RDF, although it resembles the syntax of some RDF serializations.

Web Ontology Language (OWL)

OWL is a language for creating taxonomies, ontologies, and knowledge graphs using RDF data. OWL provides vocabulary to implement the elements of an ontology, such as owl:TransitiveProperty, which specifies that a property or relation is transitive, and owl:DatatypeProperty which specifies that the range of values that a property or relation can take.

SPARQL Protocol and RDF Query Language (SPARQL)

SPARQL is a language for querying RDF data. Suppose that a supply-chain manager at a grocery store wanted to identify all of the apples in the store using a knowledge graph. A SPARQL query for getting this information would look something like the following:

SELECT ?x
WHERE {
  ?x rdf:type groceryOntology:apple
}

This query will return everything in the knowledge graph that is an instance of the class apple in the ontology used to create the knowledge graph (groceryOntology).

Advantages of Semantic Technologies

Proponents of semantic technologies claim that they have several advantages over other common data management technologies, including relational and tabular databases. These include:

• Better for integrating data from heterogeneous sources
• Better for extracting data from unstructured sources
• Easier to adapt when the data model needs to be changed
• More accessible to AI systems
• Enable automated logical inferences from existing data