GCN-based significant event trend prediction method

Abstract

The invention discloses a GCN-based significant event trend prediction method. The method comprises the following steps: acquiring structured event information data from a relational database, constructing an event semantic association graph, and selecting time slice granularity to segment a local graph; input of the GCN-based trend prediction model is node vectors and adjacency matrixes of a plurality of local event semantic association graphs, and output of the model is trend levels. According to the method, semantic understanding of the text is enhanced, and the prediction precision of the method is superior to that of a method for constructing features based on expert knowledge. The method has the advantages of being high in timeliness, high in universality and the like, and has wide application prospects in important event trend grade prediction.

Description

technical field [0001] The invention relates to the application field of natural language processing and knowledge graph, and relates to a method for constructing a semantic association graph from event information data and extracting event graph features, and extracting graph features through a graph convolution network to predict the trend level of major events . Background technique [0002] In the field of international political relations research, major events generally refer to a type of event that will have a major impact on countries or regions. Major event trend prediction divides the development trend of events into different levels, and uses the events that have occurred to predict future trend levels. At present, the main research methods can be divided into two categories: quantitative analysis based on event data analysis and machine learning classification. [0003] Traditional methods rely on expert knowledge to construct domain thematic data and quantify f...

AI Technical Summary

Problems solved by technology

At present, the method of automatic feature extraction mainly adopts the distribution feature of the subject words. This method only considers the word frequency feature, and further analyzes the news report through event extraction technology to obtain core elements such as the initiator, recipient, and event type of each event. Use event type frequency information to construct semantic and event fusion features. This method of merging event features has tried to use event data to improve document semantic understanding. When counting event type frequency information, the initiator and receiver are restricted. However, only the frequency information of high-frequency events is considered, and the event information data is not fully utilized, which splits the relationship between words or event attributes, and there is still a problem of insufficient understanding of the semantics of document features. Therefore, it is necessary to consider organizing and utilizing in a new form Event information data, document event semantic association information in rich features

Images