Research And Application Of Test Knowledge Point Classification Metho

In recent years,with the continuous introduction of national policies,the informationization construction of education is steadily advancing,and the construction of smart education is also being promoted.The classification of test question knowledge points,as a core technical support in smart education,aims to automatically predict the knowledge points tested by test questions based on the information of the test questions,and provide support for downstream tasks related to students’ learning situations(such as personalized test question recommendation,similar test question detection,etc.)by combining the students’ learning situations.Therefore,the task of test question knowledge point classification is very important.Traditional test question knowledge point classification methods often ignore the deep semantic relationship between test question images and test question texts,and do not effectively utilize the hierarchical relationship between knowledge points,resulting in insufficient understanding of test questions and poor performance of test question knowledge point classification.Therefore,this paper takes test question information as the data basis,analyzes the characteristics of test questions and knowledge points,proposes a test question knowledge point classification method based on multimodal learning and a test question knowledge point hierarchical classification method based on GCN,and designs a classification tool to apply the proposed methods in practice.The main work of this paper is as follows:(1)Test question knowledge point classification method based on multimodal learning: This method takes into account the complementary relationship between the characteristics of different modalities of test questions,and uses collaborative attention mechanism to obtain test question image features guided by test question text and test question text features guided by test question image,respectively.Then,the gate mechanism is used to dynamically fuse the features of the two to obtain richer semantic information of the test questions.Through this method,not only the classification performance of test question knowledge points can be improved,but also the feature sparsity problem in small sample test question knowledge point classification can be effectively alleviated.(2)Test question knowledge point hierarchical classification method based on GCN: Based on the previous method,this method learns the prior probability information between knowledge points,and constructs a structured encoder based on GCN using the hierarchical relationship between knowledge points,so that the fusion representation of test question text and test question image information can be propagated in this structure to strengthen the constraints between test question knowledge points.Experimental results show that the improved method can effectively improve the classification performance of test question knowledge points.(3)Design and implementation of test question knowledge point classification tool: Based on the above methods,this paper designs and implements a test question knowledge point classification tool.This tool can automatically predict the knowledge points tested by the test questions in the test paper uploaded by the user.