Methods and algorithms of neuro-fuzzy system for control of texts transfer and processing in the electronic documents

In the paper authors show constructing of systems for detection and correction of spelling mistakes on the basis of neuro-fuzzy system (NFS) methods and algorithms destined to control and correct distortions in the texts of office-work electronic documents. The ways of formalization are offered for linguistic variable and membership functions of fuzzy sets. The methods are developed for structural and parametrical identification at approximation of nonlinear dependences “input - output” of NFS. The algorithm is developed for control and correction of alarm characteristics of a text element image; it is based on synthesis of procedures such as, identification, smoothing, approximation, extraction of specific images signals characteristics from database (DB) and knowledgebase (KB). Operability of NFS is tested on the basis of Mamdani and Sugeno models in environment Fuzzy Logic Toolbox and Optimization Toolbox of MATLAB.