Work

Abusive Content Detection in Telugu-English Code-Mixed Social Media Using Hybrid Transformer Architectures

Year: 2026

Type: conference-paper

Abstract: The rapid growth of social media has led to a rise in abusive and offensive content, especially in code-mixed languages like Telugu-English.Detecting such content is challenging because of transliteration, spelling variations, informal writing styles, and frequent switching between languages.This paper presents a hybrid deep learning model for abusive content detection in Telugu-English code-mixed comments.A pretrained transformer model, DeBERTa, is used to generate contextual embeddings, which are further processed through parallel Multi-Channel CNN and BiLSTM networks for feature extraction.The CNN captures important local patterns and phrases, while the BiLSTM learns sequential and contextual information from the text.The combined features are then used for final classification.The proposed model effectively handles the linguistic complexity of Telugu-English code-mixed social media data and improves abusive content detection performance.

Source: Workshop on Speech, Vision, and Language Technologies for Dravidian Languages (DravidianLangTech)

DOI: 10.18653/v1/2026.dravidianlangtech-1.1

Authors: Bojja Revanth Reddy, Sivaiah Bellamkonda

Language: English

Topic: Hate Speech and Cyberbullying Detection

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

OpenAlex ID: W7166832886

FWCI:

Cites:

Cited by:

Related to: 10