on Artificial Neural Networks => on Artificial Neural Networks (ANNs)

ANN are => ANNs are

Conclusion in section 1.6 ends this chapter.
=>
Section 1.6 concludes this chapter.

optimilization => optimization

a output => an output

refered as => refered to as

(1.4) In this chapter -> In this section

Annex => Appendix

pattern x => pattern $\mbox{\bf x}$ etc or simply $x$

).These => ). These

values , given => values, given





minimilize => minimize (p.17)

so that can process the task more efficiently.
=> 
so that the task can be processed more efficiently.

up to what degree we want to decompose the task
=>
up to the degree to which we want to decompose the task

These concepts are quite different and complicated so we are...
=>
These concepts are quite complicated and different from each other, so we are...

informatics science => information science (or simply) informatics

The gain incorporates also => T-DTS also incorporates
(I don't understand what "gain" means)

Task decomposition makes also => Task decomposition also allows

presents theoretical basis of T-DTS approach and examples of practical 
applications and computer T-DTS implementation.
=>
presents theoretical basis of T-DTS approach, examples of its practical 
applications,  and its computer implementation.

Part concerning => The part concerning

used in part of experiments => (partly used in the experiments)

decomposition of databases => decomposing databases

5.1 that => 5.1 which

contenting => including 

As the begining part is very important to attract reader's interest,
let me ...

if my understanding/interpretation is incorrect please neglect.

====
Introduction
which is able to reduce => which can reduce (p.17)

minimilize => minimize (p.17)

Processing tasks are of very various origins.
   => Processing tasks is of very various origins. (syntactically should be)
   => Processing tasks includes verious aspects.   (semantically better) (p.17)

Sometimes there exists a relation between processing time and size of data that
makes task decomposition interesting, because of the gain of performance, possibly in
temporal or processing quality aspects. 
   => Sometimes we concern size of data we are processing. Our interest is
      decompose tasks to obtain a good performance, that is, to decrease processing time 
      and to increase processing quality, for example. (p.17)

The processing tasks we are concentrated on in this work are =>
The processing tasks we are concentrated on in this work is 

mostly model identification and classification, 
because the results are relatively easy to interpret and compare.
=>
mostly model identification for object classification, 
because this allow us relatively easy interpretation and comparison of the model. (p.17)
(Especially this depends on my understanding of your text, If it's not correct
please point it out.)

The T-DTS system [is also partially able to] attune to classification processing task by
using a family of statistical methods called Complexity Estimation techniques.
=> can also partially attune 
(syntactically --- "is able to" is usually used for a tense other than present tense)
=>The T-DTS system can also attune itself to classify processing task[s] by
using a family of statistical methods called Complexity Estimation techniques.
(semantically)

The goal is to estimate the difficulty of classification task and try to modify 
the processing algorithm to process the task efficiently.
=>
The goal is to estimate the difficulty of classification task and modify 
the processing algorithm so that it can process the task more efficiently.

The modification may include among others: task
decomposition up to some degree dependant on measurements, choice of processing
structure, choice of processing modules type for each subset of data (after decomposition).
=>
The modification may include among others: (1) What degree to which we decompose the task
depending on the mesurment of complexity; (2) How we choose the sturcture of processing; and 
(3) what processin modules' type we choose, for each subset of decomposed date.

in order to reconnaissance the task and universal and flexible
data processors (in particular Artificial Neural Networks).
=>
in order to get reconnaissance of the task and to realize 
a universal and flexible data processor in particular exploiting 
Artificial Neural Networks. (reconnaissance is not a verb but a noun)

or

in order to get reconnaissance of the task 
=> in order to well recognize the character of the task

ANN are universal as data processors because they are cross between
mathematics, statistics and informatics sciences, gaining the properties from all these
disciplines when necessary.
=>
We can think ANN  a universal data processor because it was from an intersection of 
mathematics, statistics and information science, and we can obtain the properties 
from all these disciplines when necessary.