Common basic knowledge for artificial intelligence algorithms
Completed on 05-Jul-2015 (19 days)

Project 6 in full-screenProject 6 in PDF
Basic ideas >
Data format

The proposed system is basically a big layered classification of AI (training) data, whose definition was started in the previous section by explaining the main category (i.e., Layers, associated with individual elements, as opposed to concepts). In the current section, I will be completing this definition with a detailed description of the main constituent elements of the system.

All the information accounted by this system is expected to be defined on account of the following basic elements:
  • Property
    . Bit of information (e.g., a word) suitable to be taken as a reference while comparing some other elements. Examples:
    speed
    (and associated adjectives like
    fast
    or
    slow
    ),
    size
    (and associated adjectives like
    big
    or
    small
    ) and so on.
    Properties are ranked (Ranking P) and can be defined individually (what, by bearing in mind their essence, is the most likely scenario) or on account of other elements. When being added to the system, each Property is ranked either manually (i.e., when being self-defined) or automatically (i.e., by bringing into account the rankings of its defining elements).
  • Connector
    . Complementary element required by the Properties to perform a given comparison. Examples:
    more than
    ,
    less than
    ,
    similar to
    , etc.
    Connectors are always self-defined and unranked.
  • Entity
    . It is the minimum unit of information accounted by this system which is externally relevant. It might also be understood as an individual object (e.g.,
    car
    ) as opposed to a concept (e.g.,
    "this car is fast"
    ).
    Entities are automatically ranked (Ranking E) on account of their defining elements. The definition of each Entity is always defined on account of one or more Layers (e.g., the definition of
    car
    being based upon the Layers
    Staticity (L100)
    and
    Machines (L101)
    ); additionally, it might be defined on account of one or more Properties (e.g.,
    car
    also defined on account of its
    danger to people (P0)
    and
    speed (P1)
    ).
  • Concept
    . It matches the most commonly accepted definition of concept, by bearing in mind that it will always be formed by the aforementioned elements (or/and by other Concepts).
    Equivalently to what happens with Entities, Concepts are automatically ranked on account of their defining elements. Nevertheless, this ranking has to be redefined for each single Entity, because the importance of certain concept might vary depending upon the given context. For example: the Concept
    "money is very important"
    would have a higher ranking than
    "helping others is very important"
    when analysing the Entity
    corporation
    , but not the Entity
    non-profit
    . Thus, Ranking C actually refers to a group of as many different rankings as suitable Entities (i.e., the ones having at least two associated Concepts).
    A Concept may be created from just two Entities, one Property and one Connector; but might also be built on as many Entities (+ Properties & Connectors) or other Concepts (+ Properties & Connectors) as required.

By putting together all the ideas explained in this section and in the previous one, the (ranked) layered categorisation which underlies the proposed system is formed by the following parts:
  • Two groups of elements which are ordered according to two different rankings and are externally relevant: Entities (Ranking E) & Concepts (Ranking C, which is redefined as many times as Entities with associated Concepts).
  • Two additional groups of elements which are accessorily used by the two ones in the previous point: Properties (Ranking P) & Connectors (unranked).
  • On top of all these rankings, there is another one (Ranking L) formed by enum-like elements called Layers, which defines the two aforementioned externally-relevant elements/rankings: Entities & Ranking E in a direct way; and Concepts & Ranking C indirectly (on account of the fact that Concepts are always built on Entities).
    Note that the words used to define a Layer or its (enum-)elements may also be treated as Entities or Properties. In fact, Layers are not elements of this system in a strict sense (i.e., at the same level than all the ones defined in this section); they are just a way to facilitate the understanding and usage of the proposed Data format.