Information Technology Reference
In-Depth Information
check for this. A positive:negative compound count, for example, can indicate that
the tree is possibly incorrect. If the negative count becomes too large, then the
'
tree can be split and a link added between them instead. The
new link key could then get added to some primary keysets, to allow for its tra-
versal. The trick is in being able to determine when the information is untrue. Does
the information need to be returned and then evaluated as untrue, or can the update
process allow for this to be recognised automatically? For example, a new text
sequence of
drank milk long trunk
'
is again added to the database, where it updates
the related tree nodes. As it stops short in the
'
thirsty boy drank milk
'
branch, any nodes further
up that branch can have their negative count incremented instead. As the elephant
tree reinforces the positive count here however, this is then an indication that the
tree should be split, as the information is true sometimes and false other times.
A break in any tree would automatically create a new indexing key as well. This
would be sent to all related trees that can then decide what link keyset best relates to
them
'
drank milk
'
the one for just
'
drank milk
'
or the one that also links to
'
long trunk
'
. New
entries can therefore be
first instance, until they become reliable.
There is also a process of reasoning and adjustment here, again over a period of
time. Even if a tree branch is not proven to be incorrect, in some scenarios, a
negative count might be required to indicate that part of a tree is no longer relevant
to the current entity or scenario. More traditionally, a decay factor can be used to
determine this. If, for example, the link is rarely used, its value decays until it is lost
completely. If it is used so infrequently, then it might as well not be present, even if
it is not false. So this is another alternative link update mechanism that could be
added, but a compound key helps to decide to split rather than remove completely.
With a single value that gets incremented or decremented, you have to judge how
many times each has occurred. If there is a compound count, then this is clear and it
is easy to tell if the branch is true as well as false.
fl
flagged in the
4.4 Re-join or Multiple References
A
final consideration might be the re-joining of one tree to another and also a
problem with multiple reference links to the re-joined part. As the data can be
random, it might initially be skewed in some way and force a tree to break into two
trees. Over time this evens out and the original single tree can become correct again.
This is determined by the fact that the counts become consistent with a single tree
structure again. In that case, it would be possible to re-join the previous branch that
is now a base, back onto the
first tree. The only worry would be if there are also
multiple references to the second tree that had a branch broken off. These references
might not relate to the original tree as well. It might not be good practice to allow
arbitrary references half-way up a tree and so if the previous branch has a different
set of references now, then maybe it must stay as a base. Ideas here would therefore
include transferring back only some of the new tree, while keeping the rest as the
second tree, with its base. The next section tries to explain this again, but more
Search WWH ::




Custom Search