Value Assignment Propagation (VAP)

Partitioning Classes (PTCL)

  Data Property Types (DPT):
We distinguish between four Data Property Types (DPT), namely Particular Data Properties (PDP), Transparent Data Properties (TDP), Meta Data Properties (MDP), and Universals Data Properties (UDP). The names of PDPs take a dot as their prefix, e.g. .ManufDate, while UDP names are prefixed with a combination of a dot and a circumflex, e.g. .^qtySold., MDP names start with a combination of a dot and a double circumflex, e.g., .^^multiple, and TDP names have as prefix a combination of a dot and a delta, e.g., .ΔGender.
  • The purpose of PDPs is to assert domain-specific facts exclusively to particulars (non-universals). Examples of PDPs are .DOI, .fonNr, .ISBN, .URI. Particulars cannot be instantiated (are not repeatable) from other particulars.
  • The purpose of UDPs is to assert data property values on universals (classes and object properties) that are not specific to the particulars of those universals. UDPs are derived DPs. Typically, result or range or cardinality properties such as sum, average, min, max etc. can be represented with UDPs. Examples of UDPs are .^qtySold, .^unitsInStock, .^avgLifeTime, etc. Although a UDP can be assigned different values for classes and particulars in an inheritance hierarchy, the UDP attribute-value assignments are NOT allowed to propagate down, and particulars are not allowed to instantiate UDPs.
  • The purpose of MDPs is to assert non-domain-specific metadata to knowledge objects, such as any kind of authoring and administrative information. Examples of MDPs are .^^createdBy, .^^VersionNr, .^^Particulars, .^^DevCost.
  • The purpose of TDPs is to assign values to data properties of classes where the pair (.∆attribute, value) propagates down through any entity of the hierarchy to the entities of the Particulars Layer (PL). We also refer to this method as Value Assignment Propagation (VAP). Thus, TDPs are comparable to regularity attributes as mentioned in [FoAl2021] and allow for deep instantiation. Typically, TDPs can be used to represent those properties that are characteristic of classes and their subclasses as well as all of particulars. Examples of TDPs are .∆warm-blooded, .∆ModelName, .∆BatteryPowered, .∆Gender, and so on. In our methodology, TDPs are an essential requirement for implementing the required behavior of our extended PCs.
The introduction of the four types of data properties distinguishes our approach from other Multi-Level Modeling (MLM) approaches, such as those described by Frank [Fran2014] and by Carvalho and Almeida [CaAl2016] where such a differentiated functionality is not fully available. The characteristics of the different data properties also give rise to the guidelines for selecting a specific data property type.
  Value Assignment Propagation (VAP):
The Value Assignment Propagation (VAP) method is based on the inheritance of data properties within a class hierarchy ∴c. VAP requires the following Transparent Data Properties (TDP) mechanism: Be v ∈ VSR(.Δa). If the TDP .Δa in class c has been assigned the value v with (c, .Δa, v), then this value is also propagated to all instances of all subclasses z ∈ ∴c and to all particulars >p ∈ c. Then, for any (>p, »pof, z) follows: (c, .Δa, v) ∈ KG → (>p, .Δa, v) ∈ KG and (z, .Δa, v) ∈ KG. In this sense, within the class hierarchy ∴c the VAP method extends inheritance to the inheritance of attribute-value pairs (.Δa, v) which we call features. In the next section we will formalize this with the VAP axioms PCPAx and PCCAx.

Extension: deriver.app

Back to Introduction; Deriver documentation.

Source: taoke.de — Value Assignment Propagation (VAP).