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

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, medium

Summary

Bojja Revanth Reddy and Sivaiah Bellamkonda's 2026 paper, "Abusive Content Detection in Telugu-English Code-Mixed Social Media Using Hybrid Transformer Architectures," addresses the significant challenge of identifying abusive language in low-resource, code-mixed Telugu-English social media comments. This content type is complex due to transliteration, inconsistent spelling, informal expressions, and frequent language switching within sentences. The authors propose an approach that integrates both traditional machine learning and transformer-based deep learning models. Their method involves specific preprocessing strategies to normalize transliterations and spelling variations, alongside hybrid feature extraction techniques combining TF-IDF and FastText embeddings. Furthermore, it includes fine-tuning multilingual transformer models to handle morphological complexity, contextual ambiguity, and limited annotated data, as detailed across pages 1-5 of the Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages.

Key takeaway

For NLP Engineers developing content moderation systems in low-resource, code-mixed environments, you should consider adopting hybrid transformer architectures. This approach, combining preprocessing, TF-IDF, FastText, and fine-tuned multilingual transformers, offers a robust solution for handling challenges like transliteration and limited data. Implement these strategies to improve the accuracy of abusive content detection in complex linguistic contexts, ensuring more effective platform safety.

Key insights

Hybrid transformer architectures effectively detect abusive content in challenging low-resource, code-mixed Telugu-English social media.

Principles

• Code-mixing introduces unique NLP challenges.
• Preprocessing is crucial for code-mixed text normalization.
• Combining feature types enhances detection in low-resource settings.

Method

The approach involves preprocessing for transliteration/spelling normalization, hybrid feature extraction using TF-IDF and FastText, and fine-tuning multilingual transformer models for abusive content detection.

In practice

• Apply preprocessing for code-mixed text.
• Integrate TF-IDF and FastText embeddings.
• Fine-tune multilingual transformers for specific low-resource tasks.

Topics

• Abusive Content Detection
• Code-Mixed Language
• Telugu-English NLP
• Transformer Models
• Natural Language Processing
• Low-Resource Languages
• Social Media Analysis

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• Multi-Stage Training for Abusive Comment Detection in Indic Languages — A multi-stage pipeline combining language preprocessing and model ensembles effectively detects abusive Indic language comments while minimizing false positives.
• An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect — A hybrid framework combining transformer embeddings with classical classifiers enables effective rumour detection in low-resource Algerian dialect.
• Why Nepali OCR Is Brutally Hard — And How We Finally Solved It — A hybrid deep learning pipeline overcomes complex script challenges to automate document data extraction and form pre-filling.
• Fake News Detection Using Natural Language Processing! — Combining linguistic, syntactic, and semantic features with machine learning achieves high accuracy in fake news detection.
• Building Singlish2Sinhala: A Machine Learning Approach to Sinhala Transliteration — Informal transliteration systems like Singlish present unique NLP challenges due to spelling inconsistencies.
• SenFlow: Inter-Sentence Flow Modeling for AI-Generated Text Detection in Hybrid Documents — Modeling inter-sentence dependencies via structured prediction significantly enhances AI-generated text detection in hybrid documents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.