Elamur Research⌕
☰
home / papers / 2026 / DLRG@LT-EDI 2026: Automating Counter-Narratives for Homophobic and Transphobic Comments

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

Paper · 2026 · Ramesh Kannan R and Ratnavel Rajalakshmi

Summary

The paper addresses automatic responses to homophobic and transphobic social-media comments in Tamil and English. It is a submission to Task 2 (Counter Narrative Generation) of the shared task at DravidianLangTech@LT-EDI 2026. Instead of large generative models, the DLRG team builds a lightweight pipeline from classical components:

  • A comment is first labelled as homophobic or transphobic using TF-IDF n-gram features.
  • A separate Logistic Regression model locates the harmful span.
  • A counter narrative is chosen by cosine-similarity retrieval over the top three candidates, so responses are selected rather than freely generated.

The two languages use different settings. Tamil relies on character n-grams with Multinomial Naive Bayes. English uses word n-grams, stop-word removal and a Linear SVC.

Outputs were judged with BERTScore and Distinct-2, plus rubric scores for politeness, coherence and quality. The system ranked first for Tamil and fourth for English. This suggests that cheap, retrieval-based methods remain competitive for low-resource languages. The authors acknowledge that such methods capture meaning only shallowly and limit response diversity.

Key points

  • Data: The organisers' dataset has 800 train / 109 test items for Tamil and 1800 train / 66 test items for English. Each item is labelled homophobic or transphobic.
  • Features and classifiers:
    • Tamil: TF-IDF character 3–5-grams with Multinomial Naive Bayes.
    • English: TF-IDF word 1–2-grams with Linear SVC.
    • Span detection: Logistic Regression.
    • Retrieval: cosine similarity with K=3.
  • Tamil results (1st place, 80.40% overall): BERTScore 85.73, Distinct-2 27.30, PRS 100, QS 97.71, CCNC 91.28.
  • English results (4th place, 77.29% overall): BERTScore 85.55 and Distinct-2 74.36, but weaker rubric scores (PRS 72.73, QS 69.70, CCNC 84.09).
  • Code: Available at github.com/kannanrrk/Span-Counter-Feature-Based.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Vellore Institute of Technology UniversityVellore · India7 works · 268 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 works

🧭 Topics 1Research subjects and languages the work is about.Bias, fairness & safety · 517 works

Cite

Cite as Ramesh Kannan R and Ratnavel Rajalakshmi. 2026. DLRG@LT-EDI 2026: Automating Counter-Narratives for Homophobic and Transphobic Comments. Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.ltedi-1.16

@inproceedings{r2026dlrg,
  title = {{DLRG@LT-EDI 2026: Automating Counter-Narratives for Homophobic and Transphobic Comments}},
  author = {Ramesh Kannan R and Ratnavel Rajalakshmi},
  booktitle = {Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion},
  year = {2026},
  doi = {10.18653/v1/2026.ltedi-1.16},
  url = {https://doi.org/10.18653/v1/2026.ltedi-1.16}
}

References

19 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Angana Chakraborty, Subhankar Joardar, Dilip K. Prasad, and Arif Ahmed Sekh. 2025. Graph-based hostile content detection in hindi language. Discover Computing, 28:264. Proposes a graph neural network approach combining contextual and semantic features for hostile/hate content detection in a low-reso… find
  2. Bharathi Raja Chakravarthi. 2024. Detection of homophobia and transphobia in youtube comments. International Journal of Data Science and Analytics, 18(1):49–68. find
  3. Bhargav Chhaya, Prasanna Kumar Kumaresan, Rahul Ponnusamy, and Bharathi Raja Chakravarthi. 2024. Homophobia and transphobia span identification in low-resource languages. Research on Language and Computation. find
  4. Yi-Ling Chung, Serra Sinem Tekiroğlu, and Marco Guerini. 2021. Towards knowledge-grounded counter narrative generation for hate speech. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics, pages 3846–3858. Association for Computational Linguistics. Introduces… find
  5. Anderson Almeida Firmino, Cláudio de Souza Baptista, and Anselmo Cardoso de Paiva. 2024. Improving hate speech detection using cross-lingual learning. Expert Systems with Applications, 235:121115. find
  6. Ehtesham Hashmi, Sule Yildirim Yayilgan, and Mohamed Abomhara. 2025. Metalinguist: Enhancing hate speech detection with cross-lingual meta-learning. Complex & Intelligent Systems, 11:179. Introduces a meta-learning based cross-lingual framework that combines attention mechanisms and multilingual rep… find
  7. Michael A. Hedderich, Lukas Lange, Heike Adel, Jannik Strotgen, and Dietrich Klakow. 2021. A survey on recent approaches for natural language processing in low-resource scenarios. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics, pages…
  8. Prasanna Kumar Kumaresan, Devendra Deepak Kayande, Ruba Priyadharshini, Paul Buitelaar, and Bharathi Raja Chakravarthi. 2025. Homophobia and transphobia span identification in low-resource languages. Natural Language Processing Journal, page 100169. find
  9. Durga Prasad Manukonda and Rohith Gowtham Kodali. 2024. byteLLM@LT-EDI-2024: Homophobia/transphobia detection in social media comments -custom subword tokenization with Subword2Vec and BiLSTM. In Proceedings of the Fourth Workshop on Language Technology for Equality, Diversity, Inclusion, pages 157–… find
  10. Stephen May. 2012. Linguistic Human Rights: Overcoming Linguistic Discrimination. Routledge, New York, NY. find
  11. Khouloud Mnassri, Reza Farahbakhsh, and Noel Crespi. 2024. Multilingual hate speech detection: A semisupervised generative adversarial approach. Entropy, 26(4):344. Multilingual semisupervised approach using pretrained transformers to handle data scarcity and cross-lingual detection. find
  12. Debora Nozza. 2021. Exposing the limits of zero-shot cross-lingual hate speech detection. In Annual Meeting of the Association for Computational Linguistics.
  13. Praveen Prasannan, Prasanna Kumar Kumaresan, Saranya Rajiakodi, CN Subalalitha, and Bharathi Raja Chakravarthi. 2025. Counter-speech generation for homophobic and transphobic social media content in malayalam. Social Network Analysis and Mining, 15(1):87. find
  14. Nathan Schrading, Hemant Purohit, and Amit Sheth. 2015. Analysis of persuasion techniques in tweets for public policy debates. In Proceedings of the 2015 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, pages 1206– 1213. IEEE. find
  15. Pardeep Singh, Nitin Kumar Singh, Monika, and Satish Chand. 2023. mbert-gru multilingual deep learning framework for hate speech detection in social media. Journal of Intelligent & Fuzzy Systems, 45(5):8177– 8192. Proposes a multilingual BERT (mBERT) + GRU model fine-tuned for hate speech classifica… find
  16. Kodiak R. S. Soled, Kristen D. Clark, Molly R. Altman, Jordon D. Bosse, Roy A. Thompson, Allison Squires, and Athena D. F. Sherman. 2022. Changing language, changes lives: Learning the lexicon of lgbtq+ health equity. Research in Nursing & Health, 45(6):621–632. find
  17. Muhammad Usman, Muhammad Ahmad, Irina Gelbukh, Grigori Sidorov, and Rolando Quintero Tellez. 2025. A large language model-based approach for multilingual hate speech detection on social media. Computers, 14(7):279. Hybrid framework using transformer embeddings and statistical features for multilingu… find
  18. Yifan Wang, Mayank Jobanputra, Ji-Ung Lee, Soyoung Oh, Isabel Valera, and Vera Demberg. 2025. Bridging fairness and explainability: Can input-based explanations promote fairness in hate speech detection? arXiv preprint arXiv:2509.22291. Systematic study of bias, fairness, and explainability in NLP h… — [arXiv:2509.22291](https://arxiv.org/abs/2509.22291) find
  19. Wanzheng Zhu and Suma Bhat. 2021. Generate, prune, select: A pipeline for counterspeech generation against online hate speech. In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 134–149, Online. Association for Computational Linguistics. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/r-2026-dlrg-lt-edi-2026-automating/