In order to fully utilize structured and textual patent data and realize accurate and effective disruptive technology identification,the disruptive technology identification framework based on graph neural networks is proposed with the China Patent Award as the breakthrough point.Firstly,to solve the problem of difficult technology definition,disruptive technologies are defined with the authorized invention patents that have won the China Patent Award;Then,the heterogeneous directed graph is constructed by the Neo4j graph database to store patent multiple relationships and directional information addressing the issue of low utilization of relational data;Finally,the Relationship Graph Convolutional Network(R-GCN)model is used to achieve disruptive technology identification for the solution to poor recognition effect.The research has shown that it's reasonable and reliable to directly define disruptive technologies with the patent which have won the China Patent Award;The proposed disruptive technology identification framework can make full use of patent data information and spatial information to identify the vast majority of disruptive technologies,enriching the research of graph neural networks in the area of disruptive technology identification.
维普
万方数据