An Adapted Few-Shot Prompting Technique Using ChatGPT to Advance Low-Resource Languages Understanding

📝 Overview:

We introduce a two-step approach: First, a cross-lingual transfer model generates labeled data, and second, an adapted few-shot prompting technique selects domain-specific examples to guide ChatGPT effectively. Our method enhances Intent Detection and Slot Filling performance across low-resource languages, outperforming existing techniques for Thai and Persian while achieving competitive results for Spanish.

More details on our approach will be released in this repository as soon as our paper is published!