Search papers, labs, and topics across Lattice.
The Aslema system for the NADI 2026 Shared Task 5 tackles intent recognition and slot filling by evaluating four omni LLMs in a zero-shot context against fine-tuned models, revealing that fine-tuning consistently yields superior results. Additionally, the study innovates by employing synthetic data augmentation through LLM-generated culturally grounded Tunisian Derja utterances and voice cloning, which significantly enhances task performance. Ultimately, the final model, leveraging Qwen3-Omni-30B and a blend of original and synthetic data, achieves notable accuracy metrics, ranking first in slot filling and fourth in intent recognition among competing teams.
Fine-tuning outperforms zero-shot inference, but the real game-changer is the use of synthetic data to elevate performance in culturally specific tasks.
We present Aslema, our system for NADI 2026 Shared Task 5, which consists of two subtasks: intent recognition and slot filling. We evaluate four omni LLMs in a zero-shot setting and compare them with fine-tuned models. Our results show that fine-tuning consistently outperforms zero-shot inference. We further explore synthetic data augmentation by using an LLM to generate culturally grounded Tunisian Derja utterances, followed by voice cloning to generate synthetic speech. Incorporating this synthetic data improves performance on both tasks. Our final submitted system, based on Qwen3-Omni-30B and trained with a mixture of original and synthetic data, achieves 86.8% intent accuracy and 34.7 WER on the devtest split. On the official test set it ranks 1st in slot filling (59.5 CoER) and 4th among 8 teams in intent recognition (66.1% accuracy). We release our experimental scripts and will soon share the synthetic dataset to support further research in this area.
