Research On Logistics Text Information Abstract Generation Method Based On Graph Neural Network

Logistics text information is important in logistics management.Logistics text information can link up the entire logistics system.With the advent of the era of big data technology,the logistics text information in the Internet is growing exponentially.Compressing the massive logistics text information,digging out important information from the massive logistics text information,improving reading efficiency,and solving information overload have become urgent problems to be solved.This research proposes a logistics text information summary generation method.The method uses the topic keywords as the text structure framework to guide the summary generation process.The Graph2Seq model is used to complete the summary generation task.In order to avoid excessive redundancy and repetitive information in the generated summary,the input and output sequence attention mechanism is added.The final text summary can summarize the content of the original text very well and has a certain degree of logistics professionalism.The research mainly contains two innovations.A method of keyword extraction based on graph neural network is proposed.The Graph2Seq model integrated with attention mechanism is used for logistics text summary generation.The specific work content is as follows:(1)Research on frame theory.By studing the basic theories of logistics text information summary generation method based on GNN,designed the basic model framework.(2)Subject keywords extraction.Utilizing the basic advantages of graph structure information including text semantic information,combining text vectorization and abstract semantic representation methods to generate graph structure data.Using graph neural network that aggregates and learns the characteristics of the nodes and edges in graph structure data,to learn the article content.The learned feature vector is combined with the TextRank algorithm.The TextRank value of each word is obtained through iteration.The top K words are selected as the article topic keywords.(3)A summary generation model based on the Graph2Seq model is proposed.The model in this research combines the graph structure data obtained in the keyword extraction stage and the characteristics of the graph coding structure in the Graph2Seq model.the intermediate vector in the Seq2Seq model is obtained through the node embedding and graph embedding process of the Graph2Seq model.Input and output sequence attention mechanism is combined with topic keywords and decoder generates article summary.Through experimental analysis,this paper proves that the model based on Graph2Seq combined with the subject keywords has a greater improvement in accuracy than the existing methods in the task of logistics text information summary generation.