Portable Document Format Text Analysis for Automated Question Answering

Abstract

The rapid growth of digital information has led to the storage of large volumes of textual data in PDF documents, making manual information extraction time-consuming and inefficient. This paper presents an automated question answering system for PDF text analysis using Natural Language Processing (NLP) and transformer-based models. The proposed system extracts textual data from PDF documents using the PyPDF2 library and preprocesses the extracted text through tokenization and structured formatting to match the input requirements of the T5 transformer model. The model is fine-tuned on question–context– answer datasets to generate context-aware responses to user queries. The performance of the system was evaluated using ROUGE metrics, including ROUGE-1, ROUGE-2, and ROUGE-L, which measure unigram overlap, bigram similarity, and sequence-level similarity between generated and reference answers. Experimental results show that the proposed model achieved a ROUGE-1 score of 0.5263, ROUGE-2 score of 0.3529, and ROUGE-L score of 0.5253, indicating that the system is capable of generating relevant and contextually accurate responses from PDF-based textual data.