Optimizing Question-Answering Framework Through Integration of Text Summarization Model and Third-Generation Generative Pre-Trained Transformer

This research project addresses the growing demand for efficient data access amidst the surge in digital information. Conventional keyword-based search engines face limitations, driving the exploration of advanced natural language processing (NLP) approaches. The study introduces an algorithm that autonomously extracts data from summary reports, utilizing NLP and information retrieval as a question-answering API. Evaluated using Recall-Oriented Understudy for Gisting Evaluation (ROUGE)-1, ROUGE-2, and ROUGE-L scores, PEGASUS achieved the highest average ROUGE score (0.432) with a single sample, while BART attained the highest multi-sample score (0.302) with 1000 samples. The research emphasized optimal hyperparameters in pre-trained models, specifically the impact of batch size on completion time and the relationship between maximum sequence length and ROUGE scores. The study enhances question-answering systems for efficient information retrieval, with practical applications in sectors like legal analysis, healthcare, and business intelligence. This study not only improves the efficiency and accuracy of QA systems but also offers valuable insights for future advancements in NLP-driven information extraction. The refined methodologies and enhanced performance metrics provide a promising avenue for transforming how organizations handle large-scale data, driving innovation in both computational efficiency and user experience.