Development and Performance Evaluation of Text Summarization using Deep Learning and Natural Language Processing Techniques

Every day information is produced rapidly in high volume. And finding the most relevant information from the available abundant information now a days becoming a challenge. So, to tackle these challenges now a days techiniques to extract text summary are proposed. The aim of text summarization is to produce an accurate data summary from the available vast information. The dataset we have used for doing experimental work is the Amazon Review Dataset which is publicly available on Kaggle. This research work compile that data and creates a precise text summary. A few stages of Natural Language Processing (NLP) such as Tokenization and preprocessing are used in the processing and text summary generation process. We employ the Abstractive method for text summarization; this method generates new sentences from data, some of which may not have been in the original data. We achieve this by using deep learning models, such as Long Short-Term Memory (LSTM) and Bidirectional LSTM (Bi-LSTM). The selected models trained with setting appropriate training parameters and substantial dataset. Next, the developed models’ performances are assessed using accuracy, precision, and the Bilingual Evaluation Understudy (BLEU) score as performance metrics. The results of the proposed system demonstrate satisfactory performance, achieving an accuracy of 94% and a BLEU score of 0.6.