[tcs-lc] Relationships

[Richard Pyle]

I'll take a shot at answering your questions -- not so much to give you the
answers, but to test my own understanding (and also the time-zone thing
gives me the first crack).

> The RelationshipAssertion objects are statements about something in the
> literature or contributions to the dataset, not something about logic,
> right?

Correct -- the "Assertions" part means that somebody at some time made the
subjective interpretation about the relationship between two concept
circumscriptions.  These are not deducible by logic -- only interpreted by
us feeble taxonomists.

> In particular there is nothing wrong with a concept pair having a
> RelationshipAssertion of type "is congruent to" and the same pair having
> another of type "is not congruent to" or one of type "includes" and
> another "excludes".

Correct -- with the caveat that the "AccordingTo" ought to be different for
each of the contradictory interpretations of relationship.

> If this is right, is there an expectation about
> what, if any, formal logic processing can be done on
> RelationshipAssertions? Is there an external ontology that prescribes an
> algebra on the enumeration allowed to RelationshipType?

Good question.  Certainly there is logic to the relationship types
themselves, but how a computer algorithm would deal with contradictory
RelationshipAssertions is something I've not thought about.  I know that
Nico understands this stuff very well, so perhaps he'll give you a better
answer.

> What about the corresponding questions for Relationships on Concepts?
> [They must(?) be the same answers, since both have datatype
> RelationshipType]
>
> Is there an expected connection between Relationships on a Concept and
> the RelationshipAssertions on the DataSet containing the Concept?

As I understand it (and I think Jessie has confirmed), the difference
between /TaxonConcept/Relationships and /RelationshipAssetions is that the
former are definitive, and the latter are interpretive.  All
/TaxonConcept/Relationships are "AccordingTo" the person/publication who
defined the concept (indeed, they form part of the definition itself). Given
that all /TaxonConcept/Relationships share a common AccordingTo, one would
hope that there would never be any contradictory relationships for a given
pair of Concepts.  On the other hand, RelationshipAssetions can be asserted
by anyone (specifically *not* part of the definition of the Concept), at any
time, and therefore may contain sets of contradictory assertions about the
relationships between any given pair of concepts.

As to whether or not there is a "connection" -- I imagine that relationships
in /TaxonConcept/Relationships supercede contradictory relationships in
RelationshipAssertions (i.e., definitive trumps interpretive).  Since
/TaxonConcept/Relationships *define* the concept, they cannot possibly be
considered "wrong".  But what if, in the process of comparing ones own new
concept to another (thereby creating /TaxonConcept/Relationship instances),
a well-meaning taxonomist incorrectly interpreted the concept that the new
concept is being compared to.  Does that mean that the concept
circumscription *is* something different from what the author meant?  Or
does it mean that the erroneous relationship asserted by the author of the
cocnept should be discarded from the definition, thereby preserving the
author's intended concept circumsctiption?

In any case, it seems to me that logic processing would serve a more useful
role for /TaxonConcept/Relationships, than for RelationshipAssetions.

Hope I didn't butcher that too much...

Aloha,
Rich