<p>With the rapid advancements in technology, there has been a substantial increase in the amount of information generated and shared online. This necessitates effective methods to create precise and concise document representations, allowing for faster comprehension of large amounts of data. Extractive summarization addresses this by identifying key sentences from a document to form a summary. This work introduces an innovative extractive summarization technique using an optimized deep-learning approach. The process begins by tokenizing the input document and extracting aspect terms and significant features. These are then scored using Deep Long Short-Term Memory (DLSTM). The DLSTM is a type of recurrent neural network that helps in capturing long-term dependencies within text, making it useful for scoring aspect terms and sentence relevance in the summarization process. The scores for aspect terms and features are combined through a Deep Convolutional Neural Network (DCNN), optimized using Hybrid Tuna-Whale Optimization (Hybrid TWO) to determine trainable parameters. The DCNN has the ability to capture hierarchical patterns in the text and TWO balances exploration and exploitation, which is essential for effectively tuning the parameters of the DCNN. Sentences are ranked based on their scores, and the most significant ones are combined to generate the summary. The performance of the Hybrid TWO-DCNN model is evaluated using metrics, such as Rouge-N, recall, precision, f-measure, Rouge-1, Rouge-2, and Rouge-L achieving values of 0.836, 0.936, 0.908, 0.922, 0.817, 0.799, and 0.817, respectively. The Hybrid TWO-DCNN model outperforms other state-of-the-art methods across all key metrics, highlighting its effectiveness for extractive summarization.</p>