Glossary of Ontology, Graph Database, and Knowledge Graph Terms

Organization: Topic-ordered; broad concepts precede specialized subterms.

This article aims to clarify any ambiguity regarding the terminology and conceptual definitions related to Ontology, Graph Databases, and Knowledge Graphs. As with any subject matter, achieving consensus across various fields can be challenging, as meaning is often interpreted differently depending on the context and perspective, and is therefore never universally agreed upon.

If you would like to contribute or disagree with any of the information, please send an email to km@knowledgemanagement.co.za with your suggestions or proposed updates.

1. Foundations: Ontology and Knowledge Representation

Term	Description
Ontology	A formal, explicit specification of concepts, relationships, constraints, and meaning within a domain.
Knowledge representation	The field concerned with representing facts, concepts, relationships, rules, and uncertainty in a form that computers can interpret and reason over.
Concept	An abstract idea or category represented in an ontology, such as Person, Disease, or Organization.
Class	A category whose members share common characteristics. In RDF Schema and OWL, classes describe types of entities.
Entity	A uniquely identifiable thing, real or abstract, about which information can be stated.
Individual	A particular instance of a class, such as Marie Curie as an individual of Person.
Property	A named characteristic or relationship used to connect an entity to another entity or to a value.
Object property	A relationship whose value is another entity, such as worksFor(Person, Organization).
Data property	A relationship whose value is a literal, such as a string, number, date, or Boolean.
Relationship	A meaningful connection between two or more entities. Relationships may be typed, directed, weighted, temporal, or annotated.
Semantics	The meaning assigned to symbols, terms, statements, and structures.
Syntax	The rules governing how terms and statements are written in a language or serialization.
Taxonomy	A hierarchical classification of concepts, usually organized through broader-than and narrower-than relationships.
Thesaurus	A controlled vocabulary that records synonyms, hierarchical relationships, and related terms.
Controlled vocabulary	An approved, managed set of terms used consistently for classification, indexing, or data exchange.
Conceptual model	A high-level description of important entities, concepts, relationships, and rules, independent of a particular implementation.
Formalization	The expression of domain knowledge using precisely defined symbols, structures, and logical rules.

2. Ontology Design and Structure

Term	Description
Ontology axiom	A formal statement that contributes meaning or constraints to an ontology.
Class hierarchy	An arrangement of classes according to specialization relationships.
Subclass	A class whose members are also members of a more general class.
Superclass	A more general class from which another class derives.
Subproperty	A property that is more specific than another property.
Domain	The class of entities to which a property applies.
Range	The class or datatype permitted as the value of a property.
Disjoint classes	Classes that cannot share an individual.
Equivalent classes	Classes defined as having the same members under the ontology's semantics.
Inverse property	A property that expresses the reverse direction of another property, such as hasParent and isParentOf.
Transitive property	A property for which A relates to B and B relates to C imply that A relates to C.
Symmetric property	A property for which A relates to B implies that B relates to A.
Functional property	A property that has at most one value for a given subject.
Cardinality	A constraint on the minimum, maximum, or exact number of values permitted for a property.
Restriction	A construct that limits the individuals or values associated with a class through property conditions.
Necessary condition	A condition that must be true for an individual to belong to a class.
Sufficient condition	A condition that is enough to establish class membership under the ontology's rules.
Open-world assumption	The assumption that an absent statement is not necessarily false. RDF and OWL generally use this perspective.
Closed-world assumption	The assumption that a statement not present in a database is false. Many traditional database queries use this perspective.
Unique name assumption	The assumption that different identifiers refer to different entities. RDF and OWL do not automatically make this assumption.
Ontology engineering	The process of analyzing a domain, modeling concepts and relations, formalizing them, testing the model, and maintaining it.
Ontology alignment	The identification of correspondences between concepts, properties, or individuals in different ontologies.
Ontology mapping	A set of correspondences or transformation rules relating terms in one ontology to terms in another.
Ontology modularity	The design of an ontology as separate, coherent modules that can be reused or governed independently.
Ontology versioning	The management of changes to ontology terms, identifiers, axioms, and semantics across releases.
Ontology governance	The policies, roles, review processes, and controls used to manage ontology quality and change.
Competency question	A natural-language question used to define what an ontology or knowledge graph must represent or answer.

3. RDF and Semantic-Web Data Models

Term	Description
RDF (Resource Description Framework)	A W3C data model for representing statements as subject-predicate-object triples.
RDF graph	A set of RDF triples that forms a directed, labeled graph.
Triple	An RDF statement consisting of a subject, predicate, and object.
Subject	The resource about which an RDF statement makes an assertion.
Predicate	The relationship or property asserted between the subject and object.
Object	The value of an RDF statement; it can be an IRI, blank node, or literal.
IRI (Internationalized Resource Identifier)	A global identifier that can contain a broad range of Unicode characters. RDF uses IRIs to identify resources and predicates.
URI (Uniform Resource Identifier)	A string that identifies a resource. IRIs extend URI syntax to support international characters.
Literal	An RDF value such as a string, number, date, or Boolean.
Datatype	A definition of the value space and lexical form of literals, such as `xsd:string` or `xsd:date`.
Language tag	A tag attached to a literal to identify its language, such as `en` or `fr`.
Blank node	An RDF node without a globally identifying IRI.
Named graph	An RDF graph identified by a graph name, often used for provenance, access control, or source separation.
RDF dataset	A collection consisting of a default graph and zero or more named graphs.
RDF vocabulary	A defined set of RDF terms, including classes and properties, for a domain or application.
RDFS (RDF Schema)	An RDF vocabulary and semantic extension supporting classes, subclasses, properties, domains, and ranges.
Turtle	A compact, human-readable RDF serialization that uses prefixes and abbreviated syntax.
N-Triples	A simple, line-oriented RDF serialization in which each line represents one triple.
N-Quads	An extension of N-Triples that adds a graph name for named graphs.
RDF/XML	An XML-based serialization of RDF.
JSON-LD	A JSON-based RDF serialization designed for linked-data applications.
TriG	A human-readable syntax for serializing RDF datasets, including named graphs.
RDF-star	An RDF extension that supports embedded or quoted triples, simplifying metadata about statements.
Reification	A method for making a statement itself the subject of other statements, allowing metadata about that statement to be represented.

4. OWL, Logic, and Reasoning

Term	Description
OWL (Web Ontology Language)	A W3C ontology language built on RDF that supports richer modeling and formal reasoning than RDFS.
Description logic	A family of formal knowledge-representation logics used to define OWL languages and reason about classes, properties, and individuals.
ABox	The assertion component of a knowledge base, containing facts about individuals.
TBox	The terminological component containing definitions and relationships among classes and properties.
RBox	The component containing statements about properties, such as property hierarchies and characteristics.
Entailment	A conclusion that follows logically from statements under specified semantics.
Inference	The process of deriving new statements from facts, axioms, rules, or other statements.
Reasoner	Software that computes logical consequences or checks whether conclusions follow from an ontology and data.
Inference rule	A formal pattern specifying when one or more statements imply another statement.
Forward chaining	Reasoning that starts with known facts and repeatedly applies rules to derive additional facts.
Backward chaining	Reasoning that starts with a query or goal and works backward to find supporting facts or rules.
Materialization	Precomputing and storing inferred triples so later queries can retrieve them directly.
Query-time reasoning	Computing inferences while a query is evaluated instead of storing all inferred results in advance.
Classification	Determining the class hierarchy implied by an ontology.
Realization	Determining the most specific classes to which an individual belongs.
Consistency	The property of a knowledge base whose statements and axioms do not imply a contradiction.
Satisfiability	The property of an ontology or class having at least one possible interpretation or instance.
Soundness	The property that all conclusions produced by a reasoning system are valid under the intended semantics.
OWL 2 EL	An OWL profile optimized for large class hierarchies and polynomial-time reasoning.
OWL 2 QL	An OWL profile designed for efficient query answering over large relational-style datasets.
OWL 2 RL	An OWL profile designed for rule-based reasoning that can often be implemented with rule engines.

5. Graph Data Models and Graph Databases

Term	Description
Graph data model	A data model that represents entities and relationships as connected graph elements.
Graph database	A database optimized for storing, traversing, and querying relationships among connected data.
Triple store	A database specialized for storing and querying RDF triples. It is a type of graph database.
Property graph	A graph model consisting of nodes and edges, where both can carry key-value properties.
Node	A discrete entity or vertex in a property graph.
Vertex	Another term for a node in graph theory and graph databases.
Edge	A connection between two nodes. Edges may have direction, type, weight, timestamps, or properties.
Arc	A term often used for a directed edge.
Label	A category assigned to a node or edge in a property graph.
Edge type	A classification of relationships, such as `PURCHASED`, `KNOWS`, or `LOCATED_IN`.
Path	An ordered sequence of nodes and edges connecting graph elements.
Simple path	A path in which no node is repeated.
Traversal	The process of moving through nodes and edges according to specified conditions.
Neighborhood	Nodes directly or indirectly connected to a selected node, often within a specified hop distance.
Degree	The number of edges incident to a node. In-degree and out-degree count incoming and outgoing edges.
Connected component	A maximal group of nodes connected to one another.
Graph projection	A temporary or derived graph created by selecting, transforming, or aggregating part of a larger graph.
Graph schema	The definition of permitted node types, edge types, properties, and constraints.
Schema-on-read	Applying structure and interpretation when data is queried rather than enforcing all structure at ingestion.
Schema-on-write	Requiring data to conform to a defined structure before it is stored.
Native graph database	A graph database whose storage and execution engine are designed around graph structures and traversals.
Multi-model database	A database supporting more than one model, such as graph, document, key-value, column-family, or relational.
Graph index	An auxiliary structure that accelerates lookup of nodes or edges by labels, properties, identifiers, or predicates.
Graph partitioning	Distribution of graph elements across machines or storage partitions.
Graph sharding	Horizontal partitioning in which portions of a graph are stored on different shards or servers.
ACID	Transaction properties: atomicity, consistency, isolation, and durability.
CAP theorem	A distributed-systems principle concerning trade-offs among consistency, availability, and partition tolerance.

6. Query Languages and Graph Access

Term	Description
SPARQL	The W3C query language and protocol family for querying and manipulating RDF graphs.
SPARQL query	A request that matches graph patterns and returns bindings, Boolean results, or constructed RDF graphs.
Basic graph pattern	A set of triple patterns in SPARQL that is matched against an RDF graph.
SPARQL endpoint	A network service that accepts SPARQL queries and, where supported, updates.
SPARQL SELECT	A query form that returns variable bindings in tabular form.
SPARQL CONSTRUCT	A query form that creates an RDF graph from matched data and a triple template.
SPARQL ASK	A query form that returns a Boolean indicating whether a pattern has a match.
SPARQL UPDATE	Operations for inserting, deleting, copying, moving, or modifying RDF data.
Federated query	A query that retrieves or combines data from multiple graph sources or endpoints.
Cypher	A declarative graph query language associated with property-graph systems.
Gremlin	A graph traversal language for expressing traversals over property graphs.
GQL	An ISO graph query language standard intended to provide a common language for graph data.
Graph pattern matching	Finding subgraphs that satisfy a specified structural or property pattern.
Variable-length path query	A query that matches paths within a specified length range.
Query planner	A component that selects an execution strategy for a query.
Query optimization	Improving a query plan to reduce latency, memory use, or resource consumption.
Traversal query	A query that follows edges from starting nodes according to direction, type, depth, or property conditions.

7. Knowledge Graphs and Linked Data

Term	Description
Knowledge graph	A graph-based representation of entities, relationships, attributes, and often semantic rules for integration, discovery, reasoning, and analysis.
Enterprise knowledge graph	A knowledge graph integrating an organization's systems, documents, terminology, and business concepts.
Domain knowledge graph	A knowledge graph focused on a particular subject area, such as medicine, finance, or manufacturing.
Open knowledge graph	A knowledge graph published for broad access, often under an open license or public data policy.
Knowledge graph construction	The design, population, enrichment, validation, and publication of a knowledge graph.
Knowledge graph schema	The structure governing entity types, relationship types, properties, constraints, and semantics.
Entity-centric model	A model that organizes information around identifiable entities and their relationships.
Entity resolution	Determining whether records or identifiers refer to the same real-world entity.
Entity linking	Connecting a mention in text or another data source to the corresponding entity in a knowledge base.
Entity disambiguation	Selecting the correct entity when a name or mention can refer to multiple entities.
Linked data	Structured data published using identifiers, machine-readable links, and commonly RDF-based practices.
Semantic search	Search that uses meaning, entities, relationships, ontology terms, and context rather than only keywords.
Knowledge graph completion	Predicting or discovering missing entities, relationships, or attributes.
Link prediction	Estimating whether a relationship should exist between two entities.
Knowledge graph question answering	Answering natural-language questions by interpreting them against a knowledge graph and executing queries or reasoning.
Knowledge graph fusion	Combining multiple graphs or sources while resolving duplicates, conflicts, and semantic differences.
Knowledge graph curation	Human or automated review and improvement of entities, relationships, sources, mappings, and validation results.

8. Data Ingestion, Extraction, and Integration

Term	Description
Data source	An origin of information, such as a database, API, document collection, sensor stream, or spreadsheet.
Data ingestion	Collecting and loading data into a graph or knowledge-management system.
ETL	Extract, transform, load: extracting data, transforming it, and loading it into a target repository.
ELT	Extract, load, transform: loading source data before applying transformations in the destination system.
Knowledge graph pipeline	A repeatable sequence of extraction, transformation, entity resolution, mapping, validation, loading, and monitoring.
Data transformation	Converting data from one representation, format, vocabulary, or level of granularity into another.
Schema mapping	Specifying how source fields or concepts correspond to graph classes, properties, or predicates.
R2RML	A W3C language for expressing mappings from relational databases to RDF datasets.
Direct mapping	A standardized method for deriving an RDF representation from a relational database without custom mapping rules.
Wrapper	A component that exposes source data through another interface or model without necessarily copying it.
Change data capture	Identifying and propagating inserts, updates, and deletes from a source system.
Batch loading	Loading data in discrete groups or files.
Streaming ingestion	Loading and processing graph data continuously as events become available.
Incremental update	Processing only newly added, changed, or deleted data rather than rebuilding the entire graph.
Information extraction	Automated identification of entities, facts, events, relationships, or attributes from source material.
Named-entity recognition	Identifying and categorizing mentions such as people, organizations, places, products, dates, or diseases.
Relation extraction	Identifying relationships between entities in text or other source material.
Event extraction	Identifying events, participants, properties, and time expressions.
Data integration	Combining data from different systems into a consistent view or shared representation.
Semantic integration	Integration that resolves differences in meaning, terminology, structure, and identifiers, not only differences in format.

9. Identity, Provenance, and Trust

Term	Description
Identifier	A value used to distinguish or refer to an entity, resource, statement, or record.
Persistent identifier	An identifier intended to remain stable even if a resource's location or implementation changes.
Canonical identifier	The preferred identifier selected when multiple identifiers refer to the same entity.
Same-as link	A relationship asserting that two identifiers refer to the same entity. `owl:sameAs` has strong formal semantics and should not be used merely to indicate similarity.
Provenance	Information about the origin, history, agent, method, time, or source of data and assertions.
Source attribution	Association of a statement or entity with the source from which it was obtained.
Confidence score	A numerical or categorical estimate of support for an extracted fact, entity match, link, or inference.
Trust score	An assessment of the reliability of a source, assertion, entity, or path.
Evidence	Information supporting the existence, interpretation, or validity of an entity or claim.
Claim	A meaningful assertion that may be supported, disputed, qualified, or traced to evidence.
Statement-level metadata	Metadata attached to an individual assertion, such as source, confidence, time period, or review status.
Temporal validity	The period during which a fact or relationship is considered true in the modeled domain.
Transaction time	The time at which a fact was recorded, changed, or stored.
Event time	The time at which the real-world event represented by the data occurred.
Data lineage	The trace of how data moved through systems and transformations from source to output.

10. Validation, Quality, and Governance

Term	Description
Data quality	The degree to which graph data is fit for its intended purpose. Common dimensions include accuracy, completeness, consistency, timeliness, uniqueness, and validity.
Graph quality	The quality of graph structure and content, including connectivity, identifier stability, relationship correctness, and schema conformance.
Constraint	A rule specifying what data is permitted or required, such as a datatype, cardinality, or relationship pattern.
SHACL (Shapes Constraint Language)	A W3C language for validating RDF graphs against shapes that define required properties, value types, cardinalities, and graph patterns.
Shape	A SHACL definition of constraints and validation rules applicable to a target node or graph pattern.
Validation report	A result describing whether data conforms to constraints and identifying violations or warnings.
Data profiling	Examining data to understand distributions, nulls, duplicates, value patterns, relationships, and anomalies.
Data cleansing	Detecting and correcting errors, inconsistencies, duplicates, invalid values, and formatting problems.
Deduplication	Identifying and removing or consolidating duplicate records or graph entities.
Reconciliation	Comparing data from different sources and resolving differences or selecting preferred values.
Semantic drift	A change in the meaning or usage of a term or relationship over time.
Schema drift	A change in the structure, fields, types, or constraints of a source or graph schema over time.
Data stewardship	Operational responsibility for maintaining data quality, meaning, accessibility, and appropriate use.
Data owner	The person or organization accountable for a dataset's definition, use, quality, and governance.
Access control	Policies and mechanisms determining who may read, write, update, or administer graph data.
Audit trail	A chronological record of actions, changes, approvals, or access events.
Deprecation	Formal designation that a term, identifier, mapping, or data element should no longer be used for new work.

11. Graph Analytics and Algorithms

Term	Description
Graph analytics	Algorithms and statistical methods for studying graph structure, connectivity, behavior, and patterns.
Centrality	A family of measures estimating the importance or influence of nodes.
Degree centrality	A measure based on the number of connections a node has.
Betweenness centrality	A measure based on how often a node lies on shortest paths between other nodes.
Closeness centrality	A measure based on a node's distance to other reachable nodes.
Eigenvector centrality	A measure that gives greater importance to nodes connected to other important nodes.
PageRank	A link-analysis algorithm that estimates node importance from the structure and direction of incoming links.
Community detection	Identifying densely connected groups of nodes.
Graph clustering	Grouping nodes according to structural, attribute-based, or embedding-based similarity.
Shortest path	A path between two nodes that minimizes a specified cost, such as number of edges, distance, time, or risk.
Reachability	Determining whether one node can be reached from another under specified traversal rules.
Graph pattern	A structural arrangement of nodes, edges, labels, and properties used for matching or analysis.
Motif	A small, recurring graph pattern that may represent a meaningful structural unit.
Subgraph	A graph formed from a subset of the nodes and edges of a larger graph.
Graph mining	Discovering useful patterns, associations, anomalies, or predictive relationships in graph data.
Graph neural network (GNN)	A machine-learning model that uses graph structure and node or edge features to learn representations or make predictions.
Node classification	Predicting a class or label for a graph node.
Graph classification	Predicting a class or label for an entire graph or subgraph.
Anomaly detection	Identifying unusual nodes, edges, subgraphs, values, or behaviors.

12. Emerging and Applied Concepts

Term	Description
Vector embedding	A dense numerical representation of an entity, document, term, or graph element in a multidimensional space.
Vector database	A database optimized for storing and searching numerical vectors.
Graph-vector hybrid search	Retrieval that combines graph traversal or filtering with vector similarity search.
RAG (Retrieval-Augmented Generation)	An approach in which a generative model retrieves relevant external information before producing an answer.
Graph RAG	A RAG architecture using entities, relationships, communities, or graph queries to select context for a language model.
Semantic layer	A governed layer defining shared business concepts, relationships, metrics, and meanings over underlying data sources.
Knowledge fabric	An architectural concept for connecting distributed data, metadata, semantics, and knowledge services.
Digital twin	A digital representation of a physical object, process, system, or environment, often modeled as a time-aware graph.
Event knowledge graph	A knowledge graph centered on events, participants, times, locations, causes, and consequences.
Temporal knowledge graph	A knowledge graph that explicitly represents the times associated with entities, relationships, or facts.
Probabilistic knowledge graph	A knowledge graph in which facts or relationships have probabilities or uncertainty values.
Neuro-symbolic AI	An approach combining neural machine learning with symbolic representations, rules, ontologies, or logical reasoning.
Ontology-mediated querying	Query answering in which a user queries a conceptual ontology while mappings and reasoning access underlying data sources.
Grounding	Tying an answer, claim, or generated output to identifiable source data, evidence, or graph entities.
Hallucination mitigation	Techniques that reduce unsupported generated outputs through retrieval, graph validation, provenance, citations, and checking.

13. Frequently Confused Terms

Term	Description
Ontology vs. taxonomy	An ontology can represent richer semantics, constraints, and relationship types; a taxonomy primarily organizes terms into a hierarchy.
RDF graph vs. property graph	RDF uses subject-predicate-object statements with standardized semantics. A property graph uses nodes and edges with key-value properties.
Graph database vs. knowledge graph	A graph database is primarily a storage and query technology. A knowledge graph emphasizes entities, meaning, integration, provenance, and often reasoning.
Triple store vs. graph database	A triple store is specialized for RDF. Graph database is the broader category, including triple stores and property-graph systems.
Schema vs. ontology	A schema mainly defines structure and permitted fields or relationships. An ontology additionally defines meaning and may support formal inference.
Inference vs. prediction	Inference derives conclusions that follow from rules and statements. Prediction estimates likely outcomes from learned patterns and may be uncertain.
Validation vs. reasoning	Validation checks conformance to constraints. Reasoning derives consequences from semantics and axioms.
Similarity vs. equivalence	Similarity indicates resemblance. Equivalence asserts a stronger semantic relationship.