Research On Automatic News Summarization Based On Graph Neural Networks

With the popularization and development of the internet,text information represented by news has inundated people’s lives.To obtain valuable information from massive news,people need to invest a lot of time and effort.In this context,researching automatic text summarization technology is of great significance for computers to automatically identify key information in news and generate concise and clear summaries to solve this problem.With the development of deep learning,automatic text summarization technology has achieved many excellent results,but there are still many problems,such as the fact that most of the language models currently used process text data in sequence,making it difficult to capture the relationships between sentences,especially in long documents.Most models only focus on the statistical information of sentences and words,ignoring the influence of deep semantic information on capturing important information.To address these issues,this study proposes an extractive text summarization model based on graph neural networks.The powerful representation ability of the graph data structure is used to solve the problem that existing text summarization models are difficult to model the relationships between sentences across long distances.In order to allow the model to focus on deep semantic information when identifying key information in text,this study models the relationship between sentences and implicit thematic information.The main work is as follows:1.A graph-based data structure was designed in this study,which includes three types of nodes: words,sentences,and topics.In order to construct edges between the nodes,The method takes into account their relationships,including co-occurrence and topic distribution,in order to construct edges between the nodes.2.A graph attention neural network-based automatic text summarization model is proposed in this study.3.Experimental results on the CNN/DM and NYT datasets shown that our proposed graph neural network-based automatic text summarization model outperformed baseline models,demonstrating its effectiveness for news text summarization tasks.