Elamur Research⌕
☰
home / papers / 2026 / MUCS@Dravidianlangtech@ACL2026: Hope Speech Detection in Code-Mixed Tulu Language Using Multiple Features

MUCS@Dravidianlangtech@ACL2026: Hope Speech Detection in Code-Mixed Tulu Language Using Multiple Features

Paper · 2026 · Hosahalli Lakshmaiah Shashirekha and Rachana A

Summary

The paper addresses hope speech detection in code-mixed Tulu. Tulu is an extremely low-resource Dravidian language. Online it appears in Kannada or Roman script, mixes with Kannada and English, and has no standard Romanized spelling. Team MUCS built its system for the first shared task on this topic at DravidianLangTech@ACL 2026. The task has two subtasks: coarse-grained hope tone classification (Task 1) and fine-grained hope type classification (Task 2).

No pretrained language models exist for Tulu, so the authors relied on hand-crafted features after noise-oriented pre-processing:

  • word, character and syllable n-grams;
  • BPE sub-words learned from the training set.

These features feed two soft-voting ensembles of classical classifiers: MNB with LR, and kNN with DT. An ablation study measures what each feature group contributes.

Macro F1 stays modest, roughly in the 0.4s for Task 1 and around 0.3 for Task 2. Minority classes are handled poorly because of class imbalance and overlap between categories. The work gives an early feature-based baseline for an understudied language.

Key points

  • Feature space: The combined features total 71,752 for Task 1 and 45,729 for Task 2. Syllable n-grams make up most of this: dropping them leaves 15,672 and 10,843 features respectively.
  • Accuracy vs. macro F1: In the ablation, MNB+LR reaches higher accuracy (0.66 on Task 1, 0.53 on Task 2). kNN+DT, however, appears to give better macro F1 (about 0.46–0.48 on Task 1, about 0.29–0.33 on Task 2).
  • Ablation effects: Removing a single feature group changes scores only slightly. Removing word n-grams or sub-words tends to reduce macro F1, whereas removing syllable n-grams sometimes does not hurt.
  • Weakest classes: Blended Hope and Discouraging Hope (Task 1) and Fading, Realistic and Optimistic Hope (Task 2) score worst.
  • Future work and code: The authors plan to try resampling, data augmentation and contextual or multilingual transformers. Code is released on GitHub.

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

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksPsychology · 1,649 worksSocial Psychology · 330 works

🧭 Topics 1Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 882 works

Cite

Cite as Hosahalli Lakshmaiah Shashirekha and Rachana A. 2026. MUCS@Dravidianlangtech@ACL2026: Hope Speech Detection in Code-Mixed Tulu Language Using Multiple Features. Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.dravidianlangtech-1.47

@inproceedings{shashirekha2026mucs,
  title = {{MUCS@Dravidianlangtech@ACL2026: Hope Speech Detection in Code-Mixed Tulu Language Using Multiple Features}},
  author = {Hosahalli Lakshmaiah Shashirekha and Rachana A},
  booktitle = {Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages},
  year = {2026},
  doi = {10.18653/v1/2026.dravidianlangtech-1.47},
  url = {https://doi.org/10.18653/v1/2026.dravidianlangtech-1.47}
}

References

12 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Nidhi Aggarwal, Ujjwal Singh, Venkatesh M. Reddy, and Aniket Puranik. 2023. Comparative Analysis of Machine Learning and BERT Models for Hope Speech Detection. International Journal of Social Computing. find
  2. Fazlourrahman Balouchzahi, Reza Farahbakhsh, and Noel Crespi. 2021a. Hope Speech Detection on Multilingual Social Media Texts Using Machine Learning, Deep Learning, and Transformer Models. In Proceedings of the Workshop on Online Abuse and Harms. find
  3. Fazlourrahman Balouchzahi, Reza Farahbakhsh, and Noel Crespi. 2021b. Using Ensemble Methods for Multilingual Hope Speech Detection. In Proceedings of the Forum for Information Retrieval Evaluation (FIRE 2021) – Hope Speech Detection Shared Task, pages –. ACM. find
  4. Bharathi Raja Chakravarthi. 2020. HopeEDI: A Multilingual Hope Speech Detection Dataset for Equality, Diversity, and Inclusion. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP), pages 247–256.
  5. Asha Hegde, Bharathi Raja Chakravarthi, Hosahalli Lakshmaiah Shashirekha, Rahul Ponnusamy, Subalalitha Chinnaudayar Navaneethakrishnan, Lavanya Sambath Kumar, Durairaj Thenmozhi, Martha Karunakar, Shreya Sriram, and Sarah Aymen. 2023. Findings of the Shared Task on Sentiment Analysis in Tamil and Tu… find
  6. Asha Hegde and Hosahalli Lakshmaiah Shashirekha. 2023. KT2: Kannada-Tulu Parallel Corpus Construction for Neural Machine Translation. In Proceedings of the 20th International Conference on Natural Language Processing (ICON), pages 743–753, Goa University, Goa, India. NLP Association of India (NLPAI)… find
  7. Aniket Puranik, Ujjwal Singh, and Venkatesh M. Reddy. 2023. Multilingual and Code-Mixed Hope Speech Detection using CNN and BiLSTM with BERT Variants. In Proceedings of the IEEE International Conference on Natural Language Processing. find
  8. A. Shahiki-Tash and 1 others. 2022. CNN-Based Hope Speech Detection for Spanish and English Social Media Text. Journal of Language Technology and Applications. find
  9. Poorvi Shetty. 2024. Natural Language Processing for Tulu: Challenges, Review and Future Scope. In Proceedings of the International Conference on Speech and Language Technologies for Low-Resource Languages (SPELLL 2023), volume 2046 of Communications in Computer and Information Science, pages 93–109… find
  10. Grigori Sidorov, Alexander Gelbukh, and Luis Chanona-Hernandez. 2021. Transformer-Based Models for Hope and Regret Speech Classification. In International Conference on Computational Linguistics and Intelligent Text Processing. find
  11. Tulio Ferreira Leite Da Silva, Gonzalo Freijedo Aduna, Farah Benamara, Alda Mari, Zongmin Li, Li Yue, and Jian Su. 2025. CDB: A Unified Framework for Hope Speech Detection through Counterfactual, Desire and Belief. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 4448–… find
  12. Durairaj Thenmozhi, Rathnakar Shetty P, Parameshwar R. Hegde, Anusha M D, Raksha Adyanthaya, Mohammed Fadhel Aljunid, Prasanna Kumar Kumaresan, and Bharathi Raja Chakravarthi. 2026. Findings of the Shared Task on Hope Speech Detection in Tulu. In Proceedings of the Sixth Workshop on Speech, Vision,… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/shashirekha-2026-mucs-dravidianlangtech-acl2026-hope-speech/