A NORMAL PROMPTING BASED APPROACH FOR ARABIC INTENT CLASSIFICATION

Read full text Summarise with VIRAT
Authors: fatma jamal habib, Sally Saad, Abeer Mahmoud Mahmoud
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Intent classification in natural language processing (NLP) plays a vital role in identifying users’ intentions from their spoken or written inputs. This task becomes particularly challenging in Arabic due to the language’s rich morphology and inherent ambiguity. Although recent advances in deep learning and transfer learning have significantly enhanced Arabic intent classification, a noticeable performance gap remains when compared to languages such as English.This paper introduces a novel approach to Arabic intent classification using 1-shot and 3-shot learning techniques to explore the effectiveness of a Normal prompting strategy specifically for Arabic. Experiments are conducted on a labeled Arabic intent dataset consisting of user queries covering multiple service-related intent categories, representing common real-world interactions in Arabic conversational systems. The temperature parameter is utilized to adjust probability distributions, thereby guiding the model in categorizing Arabic sentences into predefined intent classes.Experimental results indicate that prompting surpasses fine-tuning in terms of both accuracy and computational efficiency, underscoring the potential of prompting methods to advance Arabic NLP applications. The 3-shot Normal Prompt achieved an accuracy of 0.92, while fine-tuning recorded 0.87 on the same Arabic intent dataset.

Fields of Study & Subjects

Topic Modeling Domain Adaptation and Few-Shot Learning Speech and dialogue systems

Publication details

Journal / source
International journal of intelligent computing and information sciences/International Journal of Intelligent Computing and Information Sciences
Volume
26
Issue
1
Pages
149–161
Publication date
18 March 2026
Language
EN