Harnessing T5 Large Language Model for Enhanced PDF Text Comprehension and Q&A Generation

The field of Natural Language Processing (NLP) has witnessed significant advancements with the advent of transformer-based models, revolutionizing tasks such as question generation (QG) and question answering (QA). This study introduces an optimized pipeline that integrates PDF text extraction with state-of-the-art sequence to sequence models, specifically targeting the automatic generation of questions and answers from textual and PDF content. By leveraging pretrained models like T5 small and BART, proposed approach not only automates the generation of contextually relevant questions but also evaluates the performance of these models using comprehensive NLP metrics, including METEOR and BERT-Score, alongside traditional metrics like BLEU and ROUGE. proposed methodology encompasses the extraction of text from PDF documents, preprocessing and tokenization for model compatibility, and the finetuning of models on domain specific datasets to enhance performance in QG and QA tasks. This research introduces optimizations in data handling and model evaluation that significantly improve the efficiency and effectiveness of the question answer generation process. The evaluation of proposed approach using a combination of lexical and semantic metrics reveals insights into the models' capabilities in generating coherent, relevant, and semantically rich questions and answers. The findings underscore the importance of diversified evaluation metrics in capturing the nuanced performance of QG and QA models, highlighting the advantages of semantic metrics like METEOR and BERT -Score in assessing the quality of generated text. This study contributes to the NLP field by providing a robust framework for question answer generation from varied text sources, including PDFs, and by offering a comprehensive evaluation methodology that can guide future research in the area. Proposed work demonstrates the potential of integrating advanced NLP techniques to automate and enhance the question answer generation process, paving the way for applications in educational technology, content creation, and information retrieval systems.