Concept Trees: Building Dynamic Concepts from Semi-structured Data Using Nature-Inspired Methods - Complex System Modelling and Control Through Intelligent Soft Computations - page 236

(c) Even if trees have similar branches, a link might be required if other factors
do not allow a join.

5. Indexing and linking can use key sets, but it can also include graph-based
navigations. This is because the structure is tree-based, with links between tree
nodes de

ning concept sequences.

5.2 Truth Tests

1. For a tree to exist, every node in it must be true. That does not mean that every
node is evenly used, but there should be no false information. This extends to
being true for any entities that link to the tree or related sub-tree. If any part is
false for any linking entity, then the tree needs to be split.
2. Note the difference between a part of a tree that is rarely used and a part or path
that is false or untrue. Rarely used is OK, but untrue is not.
3. A set of links to a tree from different entities might make parts of the tree untrue,
when it then needs to be split at the false branch. It might however be quite
dif

cult to determine if a path is untrue, as information retrieval scenarios might
mean that the path simply does not get traversed. So the type of count key can
be important and the trick is to be able to recognise when tree paths are rarely
used, or are simply false.

(a) There could be a time-based degradation of a link, for example. If it
degrades so much as to remove it, then it has never been used and so it is
not relevant, even if it is not false.
(b) Or possibly the counting mechanism count (Greer

2011 ), that reinforces a count, both for the concept group and the indi-
vidual. This can determine when individual nodes no longer appear to be
the same as others in the group.
(c) Or there is a

' s ' group:individual '

' positive:negative ' count, when the negative count can

become too large.
(d) There could also be a time-based count that measures when events happen
at the same time. This is important for recognising when trees can be re-
joined.

5.3 Tree Comparisons

1. Tree comparisons and updates are made using groups of concepts that represent
individual input events. The event group is considered to be a complete entity
itself and gets stored in the concept base as that. It is then also compared with
the other structures as that, where it needs to match with existing tree paths in
one of two ways: