Information Technology Reference
In-Depth Information
in the middle level, then it makes sense for them to have a main base concept that
de
ne its contents. It also makes
sense for the construction process to start at the general base node and work through
to smaller and more speci
nes the tree, with branches or sub-groups that de
c details at the leaf nodes. It also makes sense that it is
more knowledge-based. The earlier neural network model (Greer 2011 ) also creates
a hierarchical structure, but it was noted that the construction process there might
start with the leaf or individual nodes that are then aggregated together into a main
or higher-level concept. That neural network model was associated more with the
third level of the cognitive model that deals more with dynamic events and triggers.
If the concept groups there are based on events, then it could make sense that a
reader of those would receive input as small events instances in time. Each event
could be some knowledge, de
ned by some structure. The events would then be
aggregated together into something more singular and maybe even learned. They
are based on time and external forces, where learning and predicting is also
important. But this then gives more sense to the architecture overall and allows for
the two hierarchical construction directions to be OK. In a general sense, we already
know this. As stated in Greer ( 2008 , Sect. 4.8), with regard to service-based net-
works: different industries would prefer either a top-down or a bottom-up approach
to organisation. Top-down starts with a central component and then adds to it when
required. Bottom-up starts with simpler components and then combines them to
provide the more complex organisation. If you want more control then a top-down
approach is preferred. If you allow a more chaotic but independent organisation,
then maybe bottom-up is preferred. It is the same argument for the cognitive model.
Top-down relates to knowledge-based concept trees and in this context also, to
small but speci
c entities. Bottom-up relates to the event-based clustering and also
to self-organising these smaller structures. As an example, you could imagine a
human seeing a tree and learning about its different components or varieties; but
when out walking on a stormy day, learning in a different way to avoid falling
branches when under a tree in high winds. Or following earlier papers
food
examples, you could imagine a human tasting different food types and learning
what they are made of; but when in a restaurant, selecting a menu based on the food
types and discovering some new recipe through the experience.
'
8 Conclusions and Future Possibilities
This paper has introduced two new ideas of concept trees and concept bases. The
concept base is a more general device that is the storage program for the trees. It is
also responsible for sorting or creating the trees, and for managing the index and link
sets. The concept trees are described in more detail and even formally de
ned. The
counting rule that is introduced in this paper and probably a different construction
method, make the concept tree a bit different to other graph-based techniques. The
addition of some rules helps to standardise the construction process and give it some
mathematical foundation. The idea of only allowing a narrowing structure with
Search WWH ::




Custom Search