Knowledge tracking method based on test question heterogeneous graph representation and learner embedding

Abstract

The invention relates to the field of education big data mining, graph neural network and learner behavior modeling, and provides a knowledge tracking method based on test question heterogeneous graph representation and learner embedding, the method adopts a heterogeneous graph neural network technology in the deep learning field to represent multi-dimensional features of test questions, and meanwhile, complex learner characteristics are modeled in combination with a project reaction theory, and ability characteristics of learners are captured by adopting a clustering mode and the like; finally, the test question and learner mixed features are fused to a traditional knowledge tracking model, and knowledge tracking and learner performance prediction for different learner groups are achieved. According to the invention, the learning condition of the learner can be scientifically and comprehensively predicted, and the purpose of assisting the teacher in precise teaching is achieved.

Description

technical field [0001] The invention relates to the fields of educational big data mining, graph neural network and learner behavior modeling, and specifically relates to a knowledge tracking method based on test question heterogeneous graph representation and learner embedding. Background technique [0002] The continuous rise of artificial intelligence and educational big data technology provides powerful technical support for precise teaching. In order to meet the requirements of precise teaching, it is necessary to study his learning trajectory from the perspective of the learner himself and predict his future learning. Response, this is what the knowledge tracking task does, that is, to model the learner's knowledge mastery status based on time, so that it can accurately track the learner's mastery of the knowledge points at the current time, and use this to predict the learner's mastery status. performance in the next learning interaction. [0003] Existing knowledge ...

AI Technical Summary

Problems solved by technology

[0006] None of the knowledge tracking methods based on matrix decomposition can obtain the dynamic change process of learners' knowledge proficiency over time. Therefore, a knowledge tracking method based on deep learning is proposed. The representative ones are the DKT model and the EKT model. The DKT model is the first The recurrent neural network is used for knowledge tracking tasks, and the LSTM model is used to track the process of learners' knowledge proficiency changing dynamically over time. However, the content of the test questions and the designed knowledge and skills still have a lot of useful information worth studying, and DKT does not. Using them effectively, EKT solves these shortcomings, but it is worth noting that EKT only uses natural language processing to process the content of the test questions, and only selects the first knowledge point involved in the test questions as a representative. It is different from the actual situation, and there is still room for improvement in the prediction accuracy, and the personal ability of the learners is also the focus of research. EKT does not use the effective information of ability

Images