Heterogeneous graph-based pre-training problem characterization method, system and device, and medium

• Main Authors: MIAO QIGUANG, HUANG SHUXUE, KWON EUI-YOUNG, RAO YU, SUN PENGGANG, SONG JIANFENG
• Format: Patent
• Language: Chinese; English
• Published: 23.06.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention discloses a pre-training problem characterization method, system and device based on a heterogeneous graph, and a medium. The method comprises the following steps: representing a relationship between whole teaching auxiliary systems by using the heterogeneous graph G; defining input information of the input layer and a processing method of the input information; performing constraint modeling by using different relationships existing in the heterogeneous graph G; the prediction layer uses the attribute features and the vertex features, performs modeling by using a PNN network, and predicts the static difficulty value of the problem; carrying out loss joint optimization to obtain characterization of a pre-training problem; according to the system, the equipment and the medium, the pre-training problem of the heterogeneous graph can be represented based on the method; according to the method, various heterogeneous information in a teaching aid system can be fully utilized, question characterizatio