Enhanced Information Retrieval Using Hybrid p-Norm Extended Boolean Models with BERT

This paper presents an Enhanced Extended Boolean Retrieval (EBR) framework that integrates Bidirectional Encoder Representations from Transformers (BERT) to address the shortcomings of traditional retrieval models, such as Term Frequency-Inverse Document Frequency (TF-IDF), which struggle with nuanced term relationships. The approach involves fine-tuning BERT and optimizing p-Norm values to enhance retrieval accuracy and query processing. Experimental results demonstrate that the Fine-Tuned BERT with EBR achieves 92% accuracy and an Area Under the Curve (AUC) of 0.92, outperforming the TF-IDF model’s 72% accuracy and traditional p-Norm models, which achieve around 83%. This hybrid framework ofers a balanced trade-of between precision and recall, proving effective for large-scale document retrieval. The study underscores the model’s scalability and robustness, highlighting its potential to enhance Information Retrieval (IR) systems for handling complex queries across various datasets.