PrimeLine@DravidianLangTech 2026: Hope Speech Detection in Tulu Using XLM-RoBERTa for Coarse and Fine-Grained Classification

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

Summary

The PrimeLine team at DravidianLangTech 2026 presented a system for hope speech detection in Tulu, a low-resource, code-mixed Dravidian language blending Tulu, Kannada script, and English. Addressing the challenge of social media content in such languages, their approach involved fine-tuning XLM-RoBERTa, a cross-lingual transformer pre-trained on over 100 languages. This was executed on task-provided datasets using Google Colab with an NVIDIA T4 GPU. The system tackled two classification tasks: a four-class coarse-grained setting (Track 1) and a five-class fine-grained setting (Track 2). On the official Codabench evaluation, it achieved a Macro F1-score of 0.34 for Track 1 and 0.19 for Track 2, establishing the first transformer-based baseline for hope speech classification in Tulu.

Key takeaway

For NLP Engineers working with low-resource or code-mixed languages, this work highlights XLM-RoBERTa as a foundational model for initial classification tasks. Given the Macro F1-scores of 0.34 and 0.19 for coarse and fine-grained hope speech detection in Tulu, you should consider this a baseline to improve upon, rather than a production-ready solution. Focus your efforts on advanced fine-tuning strategies or data augmentation to significantly enhance performance in similar challenging linguistic contexts.

Key insights

XLM-RoBERTa establishes the first transformer-based baseline for hope speech detection in low-resource, code-mixed Tulu.

Principles

• Cross-lingual models are viable for low-resource language NLP.
• Code-mixing complicates classification in social media content.

Method

Fine-tuning XLM-RoBERTa on task-specific datasets using an NVIDIA T4 GPU for multi-class hope speech classification.

In practice

• Utilize XLM-RoBERTa for initial baselines in low-resource NLP.
• Address code-mixing challenges in Dravidian language processing.

Topics

• Hope Speech Detection
• Tulu Language
• XLM-RoBERTa
• Low-Resource NLP
• Code-Mixing
• Transformer Models

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• Cross-lingual Relation Extraction with Large Language Models: Zero-Shot, Few-Shot, and Fine-Tuned Evaluation on Romanian — QLoRA fine-tuning significantly boosts cross-lingual relation extraction performance for low-resource languages, outperforming prompt-only LLMs.
• I spent 4 days training an AI model to understand my country. It barely works — Low-resource NLP projects face significant challenges in data collection, translation, and accurate labeling, often yielding modest baselines.
• 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.
• A Scalable Tool for Measuring Manner and Result Verbs in Developmental Language Research — A RoBERTa classifier, trained on LLM-generated annotations, accurately distinguishes manner and result verbs at scale.
• AfriSUD: A Dependency Treebank Collection for Evaluating Models on African Languages — Models exhibit a significant "syntax gap" in processing African languages, struggling with their structural diversity.
• 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.

Open in AIssential →

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