Abstract:Text classification is fundamental in Natural Language Processing (NLP), and the advent of Large Language Models (LLMs) has revolutionized the field. This paper introduces an adaptable and reliable text classification paradigm, which leverages LLMs as the core component to address text classification tasks. Our system simplifies the traditional text classification workflows, reducing the need for extensive preprocessing and domain-specific expertise to deliver adaptable and reliable text classification results. We evaluated the performance of several LLMs, machine learning algorithms, and neural network-based architectures on four diverse datasets. Results demonstrate that certain LLMs surpass traditional methods in sentiment analysis, spam SMS detection, and multi-label classification. Furthermore, it is shown that the system's performance can be further enhanced through few-shot or fine-tuning strategies, making the fine-tuned model the top performer across all datasets. Source code and datasets are available in this GitHub repository: <a class="link-external link-https" href="https://github.com/yeyimilk/llm-zero-shot-classifiers" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges faced by traditional text classification methods in the face of dynamic and diverse real - world environments. Specifically, traditional machine learning (ML) methods such as logistic regression and Naive Bayes have limitations in dealing with unseen data or emerging categories, and require large - scale labeled datasets, which may be difficult to obtain or impractical in practical applications. To address these issues, this paper proposes an adaptive and reliable text classification paradigm based on large - scale language models (LLMs). This method takes advantage of the powerful capabilities of LLMs, simplifies the traditional text classification workflow, and reduces the need for complex pre - processing and domain - specific expertise, thereby improving the adaptability and reliability of text classification. In addition, the system further enhances performance through few - shot learning or fine - tuning strategies, making the fine - tuned model perform well on all datasets. ### Main Problem Summary: 1. **Limitations of Traditional Methods**: Traditional machine learning methods rely on a large amount of labeled data and are difficult to adapt to unseen data or emerging categories. 2. **Reliability in Dynamic Environments**: Ensure high reliability and robustness of the model in diverse and unpredictable environments. 3. **Reducing the Technical Threshold**: Reduce the need for domain experts and complex pre - processing, enabling non - expert users to easily use high - performance classification systems. 4. **Improving Adaptability and Performance**: Enhance the performance of the model in different tasks by introducing few - shot learning and fine - tuning strategies. ### Core Contributions of the Paper: - Propose a text classification system based on LLMs, simplify the traditional classification process, and lower the technical threshold. - Introduce a new evaluation metric - Uncertainty/Error (U/E) rate, which complements the traditional accuracy and F1 - score, providing a more comprehensive evaluation of model performance. - Through experimental verification, LLMs outperform traditional methods in multiple text classification tasks after few - shot learning or fine - tuning. These improvements make LLMs more efficient, flexible, and reliable in text classification tasks, especially suitable for rapidly deployable and scalable application scenarios.

Adaptable and Reliable Text Classification using Large Language Models

Large Language Models Are Zero-Shot Text Classifiers

Mitigating Boundary Ambiguity and Inherent Bias for Text Classification in the Era of Large Language Models

Large Language Model Enhanced Machine Learning Estimators for Classification

Evaluating Large Language Models for Health-Related Text Classification Tasks with Public Social Media Data

Leveraging Label Variation in Large Language Models for Zero-Shot Text Classification

Text Clustering as Classification with LLMs

Pushing The Limit of LLM Capacity for Text Classification

Exploring LLMs as a Source of Targeted Synthetic Textual Data to Minimize High Confidence Misclassifications

Large Language Models Meet NLP: A Survey

Synthetic Data Generation with Large Language Models for Text Classification: Potential and Limitations

LLMBox: A Comprehensive Library for Large Language Models

Exploring Large Language Models for Feature Selection: A Data-centric Perspective

Sentiment Analysis in the Era of Large Language Models: A Reality Check

A Survey on Large Language Models for Recommendation

Evaluating Large Language Models: A Comprehensive Survey

Advancing Single- and Multi-task Text Classification through Large Language Model Fine-tuning

TnT-LLM: Text Mining at Scale with Large Language Models

Data Quality Enhancement on the Basis of Diversity with Large Language Models for Text Classification: Uncovered, Difficult, and Noisy

A comparative study of large language model-based zero-shot inference and task-specific supervised classification of breast cancer pathology reports