[topicmapmail] what about a "non-hierarchical chart" ?

[Ronald Poell]

Lance and Bernard

I agree with you. It is quite clear that hierarchical structures never
cover all possible views on a particular domain (may it be reduced or
extended), not even spoken about modeling our world <smile> and those of
the others</smile>.

<Lance>Could a hierarchy be considered to be no more than a selection of
useful nodes, and links between these nodes, from an n-dimensional domain
of node relationships?</Lance>.

I think you are quite right here. IMHO the use of information stored in a
network (read n-dimensional information space) depends on the context of
the user of it. If this context implicates a particular hierarchical view,
extracting this part of the network (or even a sub-network) and presenting
it to the user will be the best thing to do. Other users want different
viewpoints on the same network and seeing through their own glasses another
part of it.
Agents executing their task will need another part of the network, probably
with relationships that don't have a direct explicit meaning for a human
user. E.g. an agent will need a reference ID of a record in a particular
record in a database (e.g. ISBN number) while a human user will need
probably a textual representation of the same record (e.g. author and
title) in order to identify the record and do something with it.
For this reason I still consider that information about how to exploit
information in a network not always belongs to the information itself but
might very well be independent of that information.

If a particular relationship can be used in a transitive way it is not sure
that it should. The rule that such a relationship might be used in a
transitive way belongs to the relation, whether it effectively will be used
in that way depends on the user.

My perception of TM is that each TM presents one of the possible views on a
document (or group of documents). This viewpoint depends on the document
itself if all the information to build the TM comes from the document
itself. If the TM contains references external to the TM (like public
subjects) it participates in a conceptual network (linked through all the
available public subjects). If you associate (at some level or another)
exploitation rules (human and agent) you gradually move towards a semantic
network.
TM's are initially designed to be used as a navigation space for
information about a document. Putting a filter mechanism between the user
and the TM will produce the same effect as above. Remains the problem of
topics not related to public subjects and the related problems around
merging TM's.

This is one of the reasons why I conceived Notion System (which can be
considered close to a semantic network). But if we want to build such a
huge semantic network we will need to consider a very big picture. 

Ronald Poell
www.notionsystem.com