2.4 The differences of the cluster graphs in human brain and
computer
The acquisition of the language ability is related to the generation of
the cluster graphs. Both the morpheme graph and the signifié graph must
be generated. Human brains and computers do not necessarily process the
graphs with the same methods because of the structural differences.
There are two main possibilities for computers to express the morpheme graph: matrix and list
Adjacency list is generally unordered. By attaching the adjacency
information only to the starting nodes of the directed edges, the list
structure can be formed as a singly linked list in order to express the
directed graph. This structure makes it possible for the list to contain
the labels of each node within the list.
On the other hand, the adjacency matrix is unlikely to be suitable for
the expression of the morpheme graph. The problem is not simply the data size, but the time O(n) required to search an adjacent node
from another node: in sentence generation, this is a process to find the
subsequent morphemes.
2.5 Algorithms to produce the signifiés graph
The cluster structures can be constructed through the signifiés data
learning. The signifiés data relate arbitrary data such as video, image,
audio, temperature and so on, which can be inputted, to morphemes. AI
learns which data are related to which morphemes and vice versa. This
process is similar to the handwritten character recognition but with
larger and more various data. Similarities between several signifiés can
also be learned. For example, some “white” data can also be learned as
“bright” data and the relevance between these can be learned as well.
It is difficult to learn complexed ideas through this sort of learning.
Those complexed ideas can be learned with the help of knowledge graphs,
for example the one by Google, 2012.