Document, books summarization through Deep learning (KEY IJP************296)

• Naitik Pareek,Joel Varghese,Rishabh Gupta,Ritik Sharma, Punit Kumawat

Abstract

This project employs a deep learning-based approach in summarizing large textual material, such as books and documents. The system utilizes an encoder-decoder model with Long Short-Term Memory (LSTM) networks, along with an attention mechanism to generate abstractive summaries that resemble human-written content. The ultimate goal is to generate coherent and meaningful summaries that summarize the original work. The development was underpinned by the Agile Scrum methodology, enabling effective collaboration and tracking progress through four well-defined sprints. Preprocessing of data, model training, and testing were conducted using a range of tools, including TensorFlow, Keras, and NLTK. ROUGE and BLEU scores were used to assess the quality of generated summaries, producing promising results. The final model efficiently summarizes long inputs to concise summaries, indicating its potential application in the education, research, and publishing sectors. Future improvements include the utilization of transformer-based models and the support of multiple languages.Keywords: Deep Learning, Abstractive Summarization, LSTM, Attention Mechanism, Agile Scrum, ROUGE, BLEU, Text Summarization, Natural Language Processing, Document Summary.