A Survey of Large Language Models for Text Classification: What, Why, When, Where, and How
Zhiqiang Wang, Yanbin Lin, Jiajun Shen, Xingquan Zhu · 2025
In an age where unstructured text data is growing rapidly, effective methods for text classification(TC) have become critical. Large Language Models (LLMs), such as the revolutionary GPT-4, have taken the lead in tackling this challenge, showing remarkable abilities in handling complex language tasks. This paper presents the first thorough survey focused on LLMs for TC, a key application for managing and understanding the vast amounts of digital text we encounter today. We examine how well LLMs meet the needs of TC, explore their strengths and weaknesses, and discuss practical situations where LLMs perform best. We systematically address important questions about the effectiveness of LLMs in TC, explaining 'What' these models are in this context, 'Why' they are well-suited for these tasks, 'When' they should be used, 'Where' they have the most impact, and 'How' to use them effectively. By looking at both the benefits and challenges of using LLMs for TC, this survey aims to provide a clear guide for researchers and professionals, encouraging better use and ongoing improvements in text analysis.