Few-Shot Prompt Engineering for Multi Language Text Classification Using LlM

Abstract

Text classification is one of the important tasks in natural language processing which plays a main role in applications like sentiment analysis, spam detection, document categorization and multi-information retrieval. Traditional machine learning models require large labeled data and more feature engineering to get best accuracy. However, many real-world applications include low resource language where labeled data are scarce. This research a system for muti language text classification using few shot prompts engineering with large language models. The proposed system leverages the capability of modern large language models to understand contextual language patterns and perform classification tasks using small examples. Instead of training huge neural networks from starting, carefully designed prompts which has small number of labelled examples guide the model to perform this task effectively. The system supports multiple languages, classification text written in all the languages. An interactive interface allows the users to give multilingual text, which is processed through a prompt-based classification pipeline. The model analyzes contextual meaning, linguistic patterns, and semantic relationship to find the correct category. Experimental results shows that this few-shot prompt engineering significantly reduces the use of large data while maintain reliable accuracy