  Excuse me for my English.
	This work is devoted to decoding problem of a manuscript text by 
computer 
way. The  paper  includes some practical ideas, that must, by author opinion, 
quickly give positive results.
	At first, consider a literal data base representation  in computer
memory. It is as follow.
 1. Choice of a typical fashion of each manuscript letter.
 2. The letter is represented by coordinate matrix (fig. 1). Each cell of the
matrix corresponds to one bit in computer memory. Bit is set if a letter 
line passed through corresponding matrix cell, and is clear in other way.
After the representation have been made, we have connection between 
typical 
fashion of manuscript symbols and ASCII codes. 
For decoding it's need to read any symbol
by a scaner and transfer a part of the being read information in 
coordinate 
matrix by a software (fig 2). After this the software processing the matrix. 
The 
software treatment includes continual deformations of the being read matrix.
 Care must be taken, that the deformations is to contain gauge and angle 
transfers of matrix coordinates only. The aim of the deformations is to 
reach equality between the real matrix and any matrix from data base. Gauge
of input 
matrix for the decoder may have any value. If a decoding is fail, matrix gauge
may be varied. It allow to take in various part of being read information 
until a symbol was identified.
Base roles in this way play algoritms of matrix deformations, i haven't
finish results in the problem for the present. It is subject 
for my further search. By my opinion, the decoder must have elements of 
self learning for more effective job, that's why Prolog with predicate logic 
of first order may be useful tool. It may be applied for further treatment of
being decoded information. I've some practical results in Prolog application.
                                                                                                                                                             