Integrating particle swarm optimization with backtracking search optimization feature extraction with two-dimensional convolutional neural network and attention-based stacked bidirectional long short-term memory classifier for effective single and multi-document summarization.

The internet now offers a vast amount of information, which makes finding relevant data quite challenging. Text summarization has become a prominent and effective method towards glean important information from numerous documents. Summarization techniques are categorized into single-document and multi-document. Single-document summarization (SDS) targets on single document, whereas multi-document summarization (MDS) combines information from several sources, posing a greater challenge for researchers to create precise summaries. In the realm of automatic text summarization, advanced methods such as evolutionary algorithms, deep learning, and clustering have demonstrated promising outcomes. This study introduces an improvised Particle Swarm Optimization with Backtracking Search Optimization (PSOBSA) designed for feature extraction. For classification purpose, it recommends two-dimensional convolutional neural network (2D CNN) along with an attention-based stacked bidirectional long short-term memory (ABS-BiLSTM) model to generate new summarized sentences by analyzing entire sentences. The model's performance is assessed using datasets from DUC 2002, 2003, and 2005 for single-document summarization, and from DUC 2002, 2003, and 2005, Multi-News, and CNN/Daily Mail for multi-document summarization. It is compared against five advanced techniques: particle swarm optimization (PSO), Cat Swarm Optimization (CSO), long short-term memory (LSTM) with convolutional neural networks (LSTM-CNN), support vector regression (SVR), bee swarm algorithm (BSA), ant colony optimization (ACO) and the firefly algorithm (FFA). The evaluation metrics include ROUGE score, BLEU score, cohesion, sensitivity, positive predictive value, readability, and scenarios of best, worst, and average case performance to ensure coherence, non-redundancy, and grammatical correctness. The experimental findings demonstrate that the suggested model works better than the other summarizing techniques examined in this research.