Information Technology Reference
In-Depth Information
Fig. 1 Concept tree
generated from two text
sentences
immediate child nodes should add up to the count for the parent. If, for example, a
new list starting with
was allowed to be added, it would only increment
counts higher up the tree, altering the tree
'
sat mat
'
s balance. If this caused the triangular
rule to be broken, a re-structuring process, starting where the count becomes larger
again should
'
the tree and create a new one, with the more stable branch at
its base. As will be suggested in the rules of Sect. 5 , in fact, if the trees are always
constructed from the base up, this particular problem will not exist.
'
prune
'
4.1 Combining on Common Branches
If the following text was also stored in the concept base: The thirsty boy drank some
milk. The thirsty elephant drank some milk. This could result in two more concept
sequences:
Thirsty boy drank milk
Thirsty elephant drank milk
To add these to the concept base data structure, the process might
firstly create
two new trees, one starting with
'
thirsty boy
'
and another with
'
thirsty elephant
'
.
However, the terms
have now become the most important overall and
therefore should be at the base of a tree. Adding to Fig. 1 , the
'
drank milk
'
'
drank milk
'
branch
of the
'
black cat
'
tree should be pruned and added with the other two
'
drank milk
'
sequences, to start a new tree, with a count of 3, as shown in Fig. 2 .
It would then be necessary to add links between the trees, where they were
related. Links can be created using indexing or unique key values, for example. The
structure of each tree can then develop independently and so long as they exist, any
links between them will remain, giving some level of orderly navigation. So why
separate the concepts and not just extend the
'
black cat
'
tree? One reason is the new
base concept sets of
'
thirsty boy
'
and
'
thirsty elephant
'
, creating a new tree by
 
Search WWH ::




Custom Search