Recognition and correction of errors in the text words is necessary when words 
due to different reasons are not right. According to algorithm the system propose 
for user a short list of candidate right words ordered from the most likely 
to the least likely candidate. Binary distributed representations and methods 
of their processing (such as search and manipulations with words) were applied 
to solve this task. Experiments were made with two word sets with typical 
spelling mistakes and comparative analysis vs other methods was performed.