Overview

Original Research

BULLYING TEXT ON TAMIL-ENGLISH COMMENTS IS CLASSIFIED USING ENHANCED FEATURE EXTRACTION AND HYBRID FEATURE SELECTION TECHNIQUES

V. INDUMATHI 1, and Dr. S. SANTHANAMEGALA 2.

Vol 18, No 07 ( 2023 )   |  DOI: 10.17605/OSF.IO/MKCHU   |   Author Affiliation: Research Scholar, Assistant Professor, School of Computer Studies, Rathnavel Subramaniam College of Arts and Science, Coimbatore 1; Assistant Professor, School of Computer Studies, Rathnavel Subramaniam College of Arts and Science, Coimbatore 2.   |   Licensing: CC 4.0   |   Pg no: 1967-1976   |   Published on: 31-07-2023

Abstract

Every day, cyber bullying negative impacts on its victims get worse. Because of the harsh, emotionally abusive, and demeaning texts written by predators, several victims of cyber bullying have made suicide attempts. There have been several studies done on spotting bullying language in English comments. Bullying text categorization models are lacking in both bilingual (Tamil-English) texts, which makes the atmosphere hazardous. Due to the difficulty of classifying Bilingual texts, proposed models in Feature Extraction and Feature Selection phases which help in eliminate the Misclassification of bully texts. Proposed EW2V feature extraction model for handling the misspelled words and OOV and proposed a HPSO-GAFS hybrid model for constructing better feature sets. These models combined with the classification models to evaluate its performance. For evaluating the performance of the models Accuracy, F1-Score, Recall and Precision are used.

Keywords

Cyber Bully, Feature Extraction, Evolutionary Algorithm, PSO, GA, Feature Selection