<p>This paper presents a gated graph neural network, a general learning framework with attention to abstractive summarization of scientific documents. The problem addressed in this work is generating concise and coherent summaries of scientific documents which often contain dense and complex information. We aim to solve this by employing gated graph neural attention networks (GGNAT) which integrate graph-based neural networks with attention mechanisms to capture the text’s hierarchical relationships and contextual dependencies. This novel approach allows for a better understanding of the structure and content of scientific papers. The proposed method enhances abstractive summary accuracy and contextual awareness by delivering more precise and meaningful summaries.</p>