Interpretation of multiple values

The interpretation of an element with multiple child elements, which occur frequently in TAN files, or an attribute with multiple values can be quite unclear. Do those multiple values represent intersection, union, or distribution? For example, attribute="A B" could be interpreted to mean, using the diagram below, one instance in y (intersection), one instance in the region of x or y or z (union), or one instance in x or y and one instance in y and z (distribution).

Figure 3.1. Venn%20diagram.jpeg


The interpretation of multiple values in any TAN element or attribute is based upon perceived common usage in ordinary English language. For example, any element that takes the the section called “IRI + name Pattern” allows multiple <IRI>s. If entity j has <IRI>s A and B, and entity k has <IRI>s B and C, can j be inferred to be the same entity as k? Because people commonly use the same term while meaning different things, TAN can answer only the first half of this question. The IRI + name pattern is to be interpreted as union. But the TAN schemas cannot predict how people will interpret the extent of those two unions, or for that matter how they will interpret a single IRI.

The TAN schemas interpret the meaning of multiple values in an element or attribute in one of three ways:

Intersection. Qualifications of claims, e.g., @adverb, @claimant. For example, "...probably not..." does not mean "...probably..." and "...not..." Not a transitive property (for j = A, B; k = B, C, nothing can be inferred about the relationship between j and k).

Union (default). Anything that takes the the section called “IRI + name Pattern”, <equate-works>, @when <when>, @where. For example, "entity j is [urn:A], [urn:B]" means that entity j is urn A, urn B, or both. TAN interprets this property as being transitive (for j = A, B; k = B, C; l = C, D, one may infer j = k = l).


The interpretation of union as being transitive may result in inferences you disagree with. It is your responsibility to interrogate inferences in the TAN files you are using.

Distribution. @affects-element, @object, <object>, @src, @subject, <subject>, @verb. For example, "[Source A], [source B], are Z" means "Source A is Z" and "Source B is Z." This property is not transitive.

The above has ignored the important question of range. If entity x is said to be A, does it mean that it is true for all of x and all of A, or just some part of each? If the entity is one or more word tokens, then the statement is assumed to hold over the entire entity. If the claim is being made of a range of text, that assumption cannot be made. For example, to say that passage x quotes from passage y should not be interpreted to mean that the entirety of x quotes the entirety of y.

At present, TAN does not address this ambiguity, and leaves judgment, based on common sense, to you.