Jim> But shouldn't there be a core Person Ontology, from which to extend
Jim> or create modules?
Indeed, it is needed.
John> The design I have recommended many times is a lattice of theories.
John> Having too many axioms increases the danger of contradictions.
It is certainly a good thing to organize modules into a lattice.
A mechanism to let people and inference engines see the whole ontology
as a unique one (and derive other modules from it, using queries) is also
important for knowledge display/browsing/comparison/cross-checking purposes.
John> Every ontology is going to become a legacy system, sooner or later.
John> It is always necessary to plan for future versions, ...
Sorry if this is not very relevant in this forum but I'd like to
note that the "whole-ontology based versioning approach" (that John seems
to refer to) may not be the only solution.
There are numerous problems associated to using versions of whole ontologies
(for dealing with ontology evolution): management of identifies accross
ontologies, implicit redundancies and inconsistencies between the versions, ...
Not using "whole-ontology versions" seem avoidable when the necessary
contexts are made explicit within the ontology, for example when
- the context (time, place, author, ...) of each belief (statement that is
not a definition) is represented within the ontology, and
- each identifier for an object (category, statements, ...) includes an
identifier for a context of this object (its author, its creation date,
and sometimes some other things), e.g., as a prefix or a suffix
(this also implies that an unprefixed/unsuffixed term is an informal term,
not an identifier of an object with a unique meaning).
Indeed, doing so permits not to change identifiers and to
make explicit various beliefs according to people, time and place.
This may be seen either as a method to avoid file-based versioning (and its
numerous associated problems) or as a method to represent it in an explicit
way at a fine granularity instead of using file-based versioning.
For example, the following Formalized English sentence states that
on 31/12/2010, I (pm) asserted and believed that
- on 30/11/2009, in France, at least 50% of occurrences of
what I called birds in 2009 were able to fly, and that
- this was a correction of my belief of 30/11/2009 that
every such bird was able to fly that day in France.
pm|31/12/2010#` `at least 50% of pm|2009#bird can be agent of a flight
with place France and time 30/11/2009'
is a pm#correction of
pm|30/11/2009#`every pm|2009#fly can be agent of a flight
with place France and time 30/11/2009'
'.
This approach also permits knowledge from many sources/ontologies/modules
to be integrated in a loss-less way into a unique ontology, if only for
knowledge display/browsing/comparison/cross-checking/synthesis purposes.
This is not an argument for not also organizing modules into a lattice.
Philippe (www.phmartin.info)
