Comparative Analysis of Prompt Optimization and Fine-Tuning Techniques for Multiclass Text Classification

This thesis explores two advanced methodologies for multiclass text classification using the DBpedia dataset: prompt optimization with Stanford's DSPy framework and fine-tuning with the BERT model. The study aims to evaluate the effectiveness, efficiency, and scalability of each approach, providing insights into their applicability for large-scale natural language processing tasks. By implementing both techniques in separate modules, the research offers a comprehensive comparison of their performance and suitability for real-world applications.