NLP-Based Automatic Summarization using Bidirectional Encoder Representations from Transformers-Long Short Term Memory Hybrid Model: Enhancing Text Compression

When the amount of online text data continues to grow, the need for summarized text documents becomes increasingly important. Manually summarizing lengthy articles and determining the domain of the content is a time-consuming and tiresome process for humans. Modern technology can classify large amounts of text documents, identifying key phrases that serve as essential concepts or terms to be included in the summary. Automated text compression allows users to quickly identify the key points and generate the novel words of the document. The study introduces a NLP based hybrid approach for automatic text summarization that combines BERT-based extractive summarization with LSTM-based abstractive summarization techniques. The model aims to create concise and informative summaries. Trained on the BBC news summary dataset, a widely accepted benchmark for text summarization tasks, the model's parameters are optimized using Particle Swarm Optimization, a metaheuristic optimization technique. The hybrid model integrates BERT's extractive capabilities to identify important sentences and LSTM's abstractive abilities to generate coherent summaries, resulting in improved performance compared to individual approaches. PSO optimization enhances the model's efficiency and convergence during training. Experimental results demonstrate the evaluated accuracy scores of ROUGE 1 is 0.671428, ROUGE 2 is 0.56428 and ROUGE L is 0.671428 effectiveness of the proposed approach in enhancing text compression, producing summaries that capture the original text that minimizing redundancy and preserving key information. The study contributes to advancing text summarization tasks and highlights the potential of hybrid NLP-based models in this field.