Spell corrector to social media datasets in message filtering systems

Abstract

We develop a spell checker and corrector to check word errors in the social media datasets, which will be used in message filtering systems specially for cyberbullying detection. We use the dictionary techniques to check words and there are ten word spell error checking and correction approaches. If there are more than one corrected word we get from each approach, we use Levenshtein distance to choose the corrected word from the words in the dictionary. The spell correction results were around 90%. Moreover the percentage of each approach highlighted the efficiency of adding letters in the word.