Open AccessAn Adapted Few-Shot Prompting Technique Using ChatGPT to Advance Low-Resource Languages Understanding
Author(s) -
Saedeh Tahery,
Saeed Farzi
Publication year - 2025
Publication title -
ieee access
Language(s) - English
Resource type - Magazines
SCImago Journal Rank - 0.587
H-Index - 127
eISSN - 2169-3536
DOI - 10.1109/access.2025.3574115
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
The lack of annotated data in low-resource languages presents a significant challenge in natural language processing, particularly for language understanding tasks such as intent detection and slot filling. To address this, we propose a novel approach that first employs an effective cross-lingual transfer model to generate labeled data for the target language, overcoming the scarcity of labeled data in low-resource settings. The main contribution of our work lies in the second step, where we introduce an adapted few-shot prompting technique to guide ChatGPT as a large language model (LLM). In this step, a subset of the machine-generated examples is selected based on the domain of the input, ensuring that the LLM is provided with more tailored and domain-specific examples. This two-step process leads to enhanced performance in handling low-resource languages. We conduct extensive experiments on Spanish, Thai, and Persian using the Facebook-multilingual and Persian-ATIS datasets. Experimental results demonstrate that our method outperforms existing techniques for non-Latin languages, such as Thai and Persian, and matches state-of-the-art performance for Latin-based languages, such as Spanish. The repository for this study is publicly available at https://github.com/saedeht/language-understanding-chatgpt.