The Problem of Spelling Mistakes and Misprints Automatic Correction

Final qualifying work is dedicated to solving the problem of automatic spelling correction of words in IT-discourse texts. Texts, written by the computer specialists, include many neologisms, technicalisms and a lot of specific vocabulary. Standard spellcheckers represent low efficiency for the texts of this type. The goal of this research is to create a specialized spellchecker that copes with correcting miswritten words in IT-discourse text more efficiently than a standard spellchecker intended for texts of everyday discourse. In this research the improvements to the basic algorithm of spellcheckers have been implemented. First of all, a list of slangisms and their word forms to supplement the spellchecker dictionary has been formed. Second of all, the model that reflects the semantic connections between the words of IT-discourse and the words of everyday discourse has been trained. It was used in the spellchecker algorithm when choosing the final word to replace from the list of candidate words. Moreover, some tasks like identifying the main features of IT-discourse and determining the most productive ways of forming neologisms of IT-discourse. As a result of testing the specialized spellchecker, developed as a part of this research, the efficiency of correcting IT-discourse texts appeared to be higher than the one that «Yandex.Speller» shows by 13,3% by the F-mesure metric. This difference is explained by the ability of a specialized spellchecker to correct the mistakes in slang words and also by its ability to make choice of a final word of the candidate list more efficiently than any other spellchecker does. Improving the quality of the corrected text proved that such a way of solving the problem as creating a specialized spellchecker for a specific area of application is actual and can lead to a significant improvement in the work of automatic correction of misspelling. The main perspective of the created specialized spellchecker is its application on the site, which programmers use to discuss work issues. The ability to solve real tasks will show its real effectiveness.