Currently, the AI Knowledge-Graph is organized around 3 terminology models: the base ontology, the semantics ontology (reflects mathematical logic) and the machine-learning ontology (may be seen as informal mathematics).
The highest level contains all individuals, which are not class-identities. Class-identities are special, class-specific individuals, which can be seen as the identity of the class; there is only one class-identity per class. For example: This Dog (= class-identity) is a Dog (= class); or, in description logic notation Dog(Dog).
About 65 terms had been defined for the AI disciplines "Semantic Web Technology" and "Machine Learning" and had been automatically annotated from the internet in the 4 languages English, French, German and Spain (if available). Each explicit term/concept is defined as: Concept := SuperConcept AND distinguishing properties |Context (= terminology). Additionally, each defined concept/class contains exactly one generated class-identity.
There are about 15 generally used base terms, created in the same way as the ones above.
The lowest level contains all predicates and the relationships between individuals (currently, only between class-identities). Thus, this level contains the roles (= RBox).
Concepts
Concepts are well-defined terms, that always begin with a capital letter and may be a combined expression (for example, DescriptionLogic.
Each concept corresponds to a term defined in Wikipedia and has been interlinked to it's equivalence in DBpedia.
Each term has at least one name. If possible, generated out of DBpedia through 4 labels in English, German, French and Spanish.
Each term has at least one English description. If possible, generated out of DBpedia in English, German, French and Spanish.
Individuals and Class-Identities
Each term contains also one individual (e.g. 'Algorithm') with the same id as the corresponding concept (e.g. 'Algorithm'). In the standardized XML-formats RDF, RDFs and OWL from the World Wide Web Consortium (W3C); relationships are currently only defined for individuals and not in-between concepts. But, individuals are allowed to have the same name as a concept.
Only individuals named after the corresponding type name start with a capital letter; all other IDs of individuals start with a lowercase letter.
An individual name/ID can also been combined (e.g. 'aLetterBox').
Relationships
Relationships are defined in English.
Relationships can be compound words as well and begin always with a lowercase letter (for example, usesLogic).
Inverse relationships begin with is ... followed by the corresponding declined name (for example, isUsedBy).
Relationships take into account the well-defined characteristics of Functional, Inverse Functional, Transitive, Symmetric, Asymmetrical Reflexive and Irreflextiv.
