[tcs-lc] difference between IsParentOf/IsChildOf; and Includes/IsIncludedIn RelationshipTypes?

[Richard Pyle]

Hi Jessie,

> > 1) Does this mean that parent/child relationships should only be used
> when
> > Relationship/ToTaxonConcept points to a Concept with the *same*
> > AccordingTo
> > as the current TaxonConcept, and includes/excludes relationships
> should
> > only
> > be used when Relationship/ToTaxonConcept points to a Concept with a
> > *different* AccordingTo?  Are there any exceptions to either of these?
>
> Ok remembering there are 2 types of relationships - those defining the
> concept and those made either by the same author or by other authors
> regarding 2 concepts but not made at the main point of definition of the
> concept as published.

...the former being stored in TaxonConcept/Relationships, and the latter in
RelationshipAssertions -- right?  My intention for referencing
"Relationship/ToTaxonConcept" was to restrict it to the former (i.e.,
definitive relationships).  I probably should have been more explicit about
that.

> So for those that are defining - the AccordingTo by definition is the
> same - i.e. only those relationships stated in the AccordingTo which
> scopes the concept being defined.

Wait -- can't I define my concept by refering to concepts of other
"AccordingTo" sources? For example, suppose I want to define:

Thisgenus thisspecies Smith 2000 SEC Pyle 2005

Suppose I consider Thisgenus thisspecies Smith 2000 to be a senior synonym
of Thisgenus thatspecies Jones 2002.  Suppose further that Brown 2003
treated "thatspecies" of Jones to be a subspecies of "thisspecies".

TaxonConcepts:
A1. Thisgenus SEC Smith 2000
A2. Thisgenus SEC Jones 2002
A3. Thisgenus SEC Brown 2003
A4. Thisgenus SEC Pyle 2005
A5. T. thisspecies Smith 2000 SEC Smith 2000
A6. T. thisspecies Smith 2000 SEC Jones 2002
A7. T. thisspecies Smith 2000 SEC Brown 2003
A8. T. thisspecies Smith 2000 SEC Pyle 2005
A9. T. thatspecies Jones 2002 SEC Jones 2002
A10.T. thisspecies subsp. thatspecies Jones 2002 SEC Brown 2003

When I define my concept of T. thisspecies, can't I include *definitive*
relationships such as:

"includes T. thisspecies Smith 2000 SEC Jones 2002"
"includes T. thatspecies Jones 2002 SEC Jones 2002"

i.e., Relationships to concepts that have a different "AccordingTo"?

I already know the answer is "yes"**, so I am assuming we are suffering
miscommunication.  But the reason I spelled out the above examples is to
re-ask my original question in a different way. When I am defining
TaxonConcept A8 (T. thisspecies SEC me), would I only create an "is child
of" relationship to TaxonConcept A4 (Thisgenus SEC me)?  Or might I specify
a "is child of" definitive relationship to TaxonConcepts A1, A2, or A3?
This is what I meant by restricting Parent/Child type definitive
relationships to within-AccordingTo TaxonConcept instances.

As noted above, my definition of TaxonConcept A8 might also include
relationships such as:
"includes A6"
"includes A9"

But I also assume that I might further define my concept with:
"includes A10"

This is an example of a "Horizontal" relationship that crosses ranks --
right? (i.e., "diagonal")

Footnote:
[**User Guide,p.11: "Earlier, external concepts may be used as well to nail
down the meaning of a newly published concept."]


> For the relationship assertions, I guess people might want to assert
> parent/child relationships but what this really means is possibly
> questionable - perhaps for sparse of poorly documented concepts, someone
> might want to assert that they believe a concept is the parent or child
> of another concept - we've never said people can't do this - as from
> experience every time we tried to restrict what could be said someone
> would say they wanted to be able to do it - so we've left that up to the
> taxonomists to decide if it's something meaningful to them.

Understood.  This answers my later question on that. It's a delicate line to
walk between being restrictive and liberal in what is allowed.  Too
restrictive, we can't encode all of our data properly.  Too liberal, and
different people are expressing the same information in different ways
(defeating the purpose of an exchange schema).

> The latter - simply within and between (my simplistic mind then relates
> this to vertical and horizontal when talking with other computing folks
> - the vertical being within hierarchy and the horizontal relating two
> hierarchies via some node)

O.K., thanks -- so "vertical" and "horizontal" are not defined in terms of
taxonomic Rank.  Got it.

> > I don't necesarily agree we need both (pending the answer to my
> question
> > #1
> > above), except as a performance enhancement device to save the
> processing
> > time required to compare "AccordingTo"s of two TaxonConcet instances
> > linked
> > by a Relationship of Type "includes/included in".  This is complicated
> > further by the "has synonym" RelationshipType.  In any case, clearly a
> > point
> > of discussion that exceeds the definition of "minor".
>
> Well I don't know now whether you think we need both or not

I'm still not sure, because I don't yet know the answer to my question #1.
But generally, my feeling about Concept-Concept relationships is that they
should be completely independent of taxonomic Rank -- i.e., they should only
relate to set-theory relationships, and not confound things by using
taxonomic hierarchy to artificially create a relationship type that doesn't
convey any different logical information from another relationship type.  As
far as I can see, "A1 Is Parent of A2" doesn't tell us anything different
from "A1 Includes A2", except that the the TaxonNames attached to each of
these two concepts are at different ranks (and, perhaps, also that they
share an AccordingTo --depending on the answer to my Question #1). This is
something that is already self-evident from data already associated with
these two TaxonConcept, so the only advantage (as far as I can see) of
allowing both types of Relationships is to save the trouble of looking up
the Rank and AccordingTo associated with two related TaxonConcepts.  A very
small advantage, by my estimate.

> - just
> depends on the level of semantics to be captured - I still think we need
> both - but no-one is forcing you to use both if you don't want to or if
> you believe there will be no loss of semantics by using only one.

Understood.  I guess it's something we can test with real-world data sets.
But again, if we allow too much "personal preference" in how data are
represented, we may start losing the value of having a universally adopted
exchange schema in the first place.

Aloha,
Rich

Richard L. Pyle, PhD
Ichthyology, Bishop Museum
1525 Bernice St., Honolulu, HI 96817
Ph: (808)848-4115, Fax: (808)847-8252
email: deepreef at bishopmuseum.org
http://hbs.bishopmuseum.org/staff/pylerichard.html