Bi-STAB: Enhanced abstractive text summarization model using hybridized deep learning network

Abstractive text summarization is based on Sequence-to-Sequence architecture which is developed to capture syntactic and contextual information between words. Abstractive summarization is needed to address information overload and enhance summary generation. However, it faces challenges such as maintaining consistency, capturing subtle nuances, and striking a balance between conciseness and comprehensiveness. To overcome these challenges, a novel Bi-directional Stacked GRU-LSTM for ABstractive text summarization (Bi-STAB) framework is proposed to generate an effective text summarization system using a hybridized deep learning network. Initially, the original document is pre-processed and fed to the hybridized stacked deep learning network for relevant feature extraction. The hybridized deep learning network extracts features from the text by capturing both forward and backward sequential dependencies by ensuring contextual information from both directions. An Attention-based Neural Network is used to select the most relevant parts of the text to form a coherent summary from these extracted features. Finally, the summary is then post-processed to enhance readability and provide a concise and accurate representation of the original text. A simulation of Bi-STAB is performed using MATLAB, and validation is performed using WCEP and Giga Word data. TBy comparing the accuracy of Bi-STAB with existing techniques such as DeepSumm, MOOTweetSumm, and WL-AttenSumm the proposed framework outperforms these techniques by 84.80, 88.35, 92.17, and 97.64%, respectively.