Double convolution knowledge tracking method and system integrating question mode and answering results

Abstract

The invention discloses a double convolution knowledge tracking method and system that integrates question patterns and answer results. The method includes the steps of: obtaining the historical answer data set of the answerers; extracting each answerer's question sequence, answer result sequence and skill sequence from the historical answer data set; splicing the question sequence and skill sequence and inputting them into the first dimension The convolutional neural network extracts the characteristic data of the question pattern; the answer sequence is input to the second one-dimensional convolutional neural network to extract the characteristic data of the answer result; the characteristic data of the question pattern and the characteristic data of the answer result are spliced ​​and then input to the fully connected layer The network outputs the predicted result data of the respondent's answering behavior. The present invention extracts the characteristic data of the question pattern and the characteristic data of the answer result through modeling, and integrates these two characteristics to carry out knowledge tracking, which can improve the prediction accuracy.

Description

technical field [0001] The invention belongs to the technical field of knowledge tracking, and more specifically relates to a double convolution knowledge tracking method and system that integrates question patterns and answer results. Background technique [0002] Knowledge tracking is to use computer technology to model the knowledge state of the answerer, so as to track the answerer's mastery of the knowledge points, and further predict the answerer's performance in the next answer. Knowledge tracking can capture the real situation of the answerer currently doing the question, and is the core task in learner modeling. Knowledge tracking is widely used in the field of intelligent education, for example, learning resources can be automatically recommended based on knowledge tracking predictions. [0003] There are several classic models in the field of knowledge tracking, such as Deep Knowledge Tracing (DKT), Dynamic Key-Value Memory Networks (DKVMN), Convolutional Knowled...

AI Technical Summary

Problems solved by technology

[0004] However, knowledge tracking using the CKT model also has the following problems: every time the answerer does a question, the state of knowledge mastery of the answerer will change. Performance when answering questions

On the other hand, CKT uses the learning rate to express it roughly, but does not make full use of the characteristics of each question, and does not consider the dynamic changes in the knowledge mastery state of the respondent during the answering process. In terms of prediction accuracy, the area under the curve (AUC) of CKT The value is 0.822, and there is still room for improvement in prediction accuracy

Images