In an era of rapidly growing information, efficient text summarization is key for extracting key visions. The paper presents a method for generating summary from articles using a Pointer-Generator Network (PGN) combined with attention distribution and coverage mechanisms. The PGN addresses issues like repetition and inaccuracy by combination of extractive and abstractive summarization techniques. The model, trained on the CNN/DailyMail dataset, utilizes tokenization, padding, and advanced optimization methods. PGN outperforms the Bidirectional LSTM (BD-LSTM) across all ROUGE metrics. Specifically, PGN achieves a higher ROUGE-1 score of 38.86%, compared to BD-LSTM 30.70%, indicating better unigram overlap. PGN also scores 16.70% in ROUGE-2, surpassing BD-LSTM 11.67%, showing improved bigram overlap. In ROUGE-L, PGN scores 35.38%, significantly outperforming BD-LSTM 24.97%. These results highlight PGN superior ability to generate more coherent and accurate summaries, making it a more effective model for automatic text summarization.