Korean Malicious Text Analysis Technique Applying Consonant and Vowel Separation Method

Abstract

Social costs due to malicious writing, including chatting and commenting, are increasing. There are many techniques to filtering malicious writing, but there are problems that make it inefficient because of detection bypass techniques such as typos and misspellings, which make detection avoidable. Since these sentences cause problems even in the preprocessing process through morpheme analysis a preprocessing technique was proposed that applies a particle that separates consonants and vowels. Among the various methods in this paper, the first consonant, middle consonant, final consonant, and non-complete Hangul consonant form that do not include Korean spaces showed the best performance with an accuracy of about 78%.