Team Oryu@DravidianLangTech 2026: A Multilingual Transformer Approach for Hope Speech Detection in Code-Mixed Tulu

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

Summary

Team Oryu submitted a system for Task 1: Coarse-Grained Hope Tone Classification in Code-Mixed Tulu at DravidianLangTech 2026, aiming to foster positive communication on social media. Their system classifies social media comments into four categories: Encouraging, Discouraging, Uninvolved, and Blended Tone. Addressing the challenge of heavy code-mixing among Tulu, English, and Kannada, the team employed a fine-tuned multilingual transformer model, alongside code-mixed text processing, data augmentation, and class-weighted loss to manage class imbalance. This approach secured 3rd position in the shared task with a Macro F1-score of 63%. The results highlight the effectiveness of multilingual transformer models for emotionally nuanced classification in code-mixed environments, while also pointing to the inherent difficulties in accurately capturing "Blended Tone" classifications.

Key takeaway

For NLP Engineers developing social media sentiment analysis in low-resource, code-mixed environments, consider fine-tuning multilingual transformer models. Your approach should integrate code-mixed text processing, data augmentation, and class-weighted loss to improve performance and address data imbalance. Be aware that accurately classifying "Blended Tone" remains a complex challenge, requiring further research or specialized handling in your system design.

Key insights

Multilingual transformers effectively classify hope speech in code-mixed low-resource languages, despite challenges with blended emotional tones.

Principles

• Multilingual transformers excel in code-mixed emotional classification.
• Data augmentation and class-weighted loss enhance model robustness.
• Blended emotional tones pose a significant classification challenge.

Method

The system fine-tuned a multilingual transformer model, applied code-mixed text processing, utilized data augmentation, and incorporated class-weighted loss to handle class imbalance for hope speech detection.

In practice

• Apply multilingual transformers for low-resource language tasks.
• Use data augmentation for code-mixed text.
• Implement class-weighted loss for imbalanced datasets.

Topics

• Hope Speech Detection
• Multilingual Transformers
• Code-Mixed NLP
• Tulu Language
• Low-Resource Languages
• Sentiment Classification

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• mdok-style at SemEval-2026 Task 9: Finetuning LLMs for Multilingual Polarization Detection — Finetuning mid-sized LLMs with QLoRA and augmented data improves multilingual online polarization detection.
• TeamHerald@CHIPSAL 2026: Hate Speech Detection and Sentiment Analysis of Nepali Memes using Transformer-based Architectures and Ensemble Learning — Ensemble strategies for meme analysis perform differently for binary vs. multi-class tasks, requiring task-specific aggregation.
• Comparison of Modern Multilingual Text Embedding Techniques for Hate Speech Detection Task — Supervised models with multilingual embeddings and gradient boosted trees excel at hate speech detection.
• Turkish Sentiment Analysis (Keras LSTM + BERT) — A hybrid NLP pipeline combines custom LSTMs and pre-trained BERT for robust Turkish sentiment analysis with real-time data acquisition.
• Indigenous Letters to Brazil: Multi-Label Classification — Multi-label classification of indigenous letters presents challenges due to corpus imbalance and semantic overlap.
• Low-resource Language Discrimination Towards Chinese Dialects with Transfer learning and Data Augmentation — Transfer learning and data augmentation effectively overcome low-resource challenges in Chinese dialect discrimination.

Open in AIssential →

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