The Use of Dynamic n-Gram to Enhance TF-IDF Features Extraction for Bahasa Indonesia Cyberbullying Classification

Cyberbullying detection in a sentence or utterance has challenges due to syntactic and meaning variations (lexical). Term Frequency-Inverse Document Frequency (TF-IDF) carries out textual feature extraction to produce candidates thematically based on word occurrence statistics. However, these candidates are generated without considering a term relationship between constituent elements in the parsing language syntax. This study discusses a TF-IDF feature extraction model using the n-Gram approach to produce candidate feature selection based on a specified term relationship. Thresholding applications for the formation of dynamic n-Gram segmentation were also discussed. Furthermore, the dynamic n-Gram model in TF-IDF feature extraction can be used in cyberbullying classification to overcome variations in syntax and meaning of sentences/speech from Bahasa Indonesia.