Information Technology Reference
In-Depth Information
In data analysis processes run by cognitive information systems, a certain fea-
ture characteristic for these very systems is noted. Namely, data analysis processes
cannot end at the stage of recognising the analysed data sets, because the essence
of the process conducted is to extract the features characteristic for this data and
determine their semantics. This is why the processes of automatic data analysis
now include a stage of automatic understanding executed using artificial intelli-
gence technologies, which, apart from the simple recognition of the item to be
analysed, can also extract significant semantic information from the set. This in-
formation allows its meaning to be interpreted, in other words, supports its full
understanding.
The system learning process progresses in five stages. At the first stage, the
semantic information which can be of significance in the further analysis process
is extracted from sets of solutions, so if some information is useless in the current
data analysis process, it is treated as superfluous for that process and will be omit-
ted from it. After this stage, the features characteristic for the solution obtained at
the first stage of the analysis process are identified. Identifying that type of fea-
tures can cause a change of the solution obtained at the first stage, e.g. because the
set of expectations or the expert knowledge base is extended and thus new patterns
are defined. This moment marks the beginning of the following stage which fo-
cuses on indicating significant changes in the field of characteristic features lead-
ing to improving (optimising) the solution formulated, as a result of which, a re-
definition is carried out in the set of characteristic features in the system. The last
stage of system learning is that of looking for solutions based on a new set of
characteristic features and a new expert knowledge base, which at this stage takes
the shape of a set of new patterns defined in the cognitive system.
Supplementing data analysis with stages at which the system learns new solu-
tions means that the cognitive resonance must be repeated in the data analysis
process, and if the learning process is multiplied, then cognitive resonance must
be repeated more than once. Incorporating new system learning solutions in the
data analysis process makes those data analysis processes much more complex
(Fig. 2.12).
After the stage at which cognitive systems learn new solutions as part of data
analysis processes, the analysis and understanding process based on cognitive
resonance is repeated, but unlike the traditional analysis process, is now makes use
of new (extended) sets of analysed data and a new (extended) base of expert
knowledge. It is these very elements that become the primary foundation of cogni-
tive data analysis processes in new analysis and understanding systems enhanced
with aspects of cognitive system learning.
