DLRG@LT-EDI 2026: Automating Counter-Narratives for Homophobic and Transphobic Comments

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cybersecurity & Data Privacy · Depth: Expert, quick

Summary

A computational framework for generating Counter Narratives (CNs) against online homophobic and transphobic hate speech, particularly in low-resource languages like Tamil, has been proposed. It uses classical NLP techniques, specifically TF-IDF features with n-grams, for identifying homophobic or transphobic labels and performing span detection. Counter narratives are retrieved by computing cosine similarity, ensuring semantic alignment and contextual relevance. The system was evaluated on an expanded human-curated dataset, achieving an overall average score of 80.40 % for Tamil and 77.29 % for English in Task 2 of LT-EDI 2026, securing first and fourth rank respectively. This framework addresses the challenge of effective content moderation where automated systems are lacking.

Key takeaway

For NLP Engineers developing content moderation systems, you should consider integrating classical NLP techniques like TF-IDF and cosine similarity for automated counter-narrative generation. This approach effectively addresses hate speech in low-resource languages, as demonstrated by its 80.40 % score for Tamil. You can explore expanding human-curated datasets to improve system performance and contextual relevance, particularly for specific hate speech categories like homophobia and transphobia.

Key insights

A computational framework automates counter-narrative generation for homophobic and transphobic online hate speech using classical NLP.

Principles

• Classical NLP can combat hate speech.
• Semantic alignment improves CN relevance.
• Low-resource languages need specific solutions.

Method

The framework uses TF-IDF features with n-grams for label identification and span detection. Counter narratives are retrieved via cosine similarity for semantic and contextual relevance.

In practice

• Apply TF-IDF for hate speech classification.
• Use cosine similarity for CN retrieval.
• Expand human-curated datasets.

Topics

• Counter-Narrative Generation
• Hate Speech Detection
• TF-IDF
• Cosine Similarity
• Low-Resource NLP
• Content Moderation

Code references

• kannanrrk/Span-Counter-Feature-Based

Best for: Research Scientist, AI Scientist, NLP Engineer, AI Ethicist

Related on AIssential

• Assisted Counterspeech Writing at the Crossroads of Hate Speech and Misinformation — LLMs can assist counterspeech against co-occurring hate and misinformation, but expert refinement and mixed knowledge strategies are crucial.
• Tokenization’dan TF-IDF’e: Metin Verisini Makine Öğrenmesine Hazırlamak — This content outlines the fundamental steps for preparing raw text data for machine learning models within an NLP pipeline.
• Aligning Implied Statements for Implicit Hate Speech Generalizability with Context-Bounded Semi-hard Negative Mining — The ImpSH framework improves implicit hate speech detection by aligning posts with implied statements and using context-bounded semi-hard negatives.
• Explain the Flag: Contextualizing Hate Speech Beyond Censorship — A hybrid LLM and vocabulary approach improves hate speech detection and explanation across multiple languages.
• The Cosine Similarity Trap: Why Embeddings Can’t Distinguish “War” from “Union” — Language models often cannot distinguish complementarity from conflict due to how embeddings encode relatedness, not relationship.
• Socially Responsible and Explainable Automated Fact-Checking and Hate Speech Detection — Explainable NLP methods and high-quality datasets can jointly optimize performance and interpretability for Portuguese hate speech and fact-checking.

Open in AIssential →

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