%0 Conference Paper %1 10.1007/978-3-031-30105-6_19 %A Chen, Jisen %A Shen, Jian %A Long, Ting %A Shen, Liping %A Zhang, Weinan %A Yu, Yong %B Neural Information Processing %C Cham %D 2023 %E Tanveer, Mohammad %E Agarwal, Sonali %E Ozawa, Seiichi %E Ekbal, Asif %E Jatowt, Adam %I Springer International Publishing %K knowledge-tracing neural-network %P 224--235 %T Heterogeneous Graph Representation for Knowledge Tracing %X Knowledge tracing (KT) is a fundamental task of intelligent education, which traces students' knowledge states by their historical interactions. In KT, students, questions, concepts, and answers are four main types of entities, and they contain various relations, including student-question interactive relations, question-concept relations, and question-answer relations. Such rich knowledge in these heterogeneous relations could potentially improve the prediction of KT. However, it has not been sufficiently utilized in existing KT methods. In this paper, we propose a novel method, called Heterogeneous Graph Representation for Knowledge Tracing, to leverage these useful relations. Our method first models all the complex entities and relations in KT as a heterogeneous graph, and then uses a heterogeneous graph neural network to obtain entities' feature representations. After that, we feed the feature embeddings to a KT model in an end-to-end training manner. Due to the heterogeneous graph's high representational capacity, our method exploits the relations among students, questions, concepts, and answers in a concise and unified way. Experiments on four KT datasets show that our method achieves state-of-the-art performance. %@ 978-3-031-30105-6

@inproceedings{10.1007/978-3-031-30105-6_19, abstract = {Knowledge tracing (KT) is a fundamental task of intelligent education, which traces students' knowledge states by their historical interactions. In KT, students, questions, concepts, and answers are four main types of entities, and they contain various relations, including student-question interactive relations, question-concept relations, and question-answer relations. Such rich knowledge in these heterogeneous relations could potentially improve the prediction of KT. However, it has not been sufficiently utilized in existing KT methods. In this paper, we propose a novel method, called Heterogeneous Graph Representation for Knowledge Tracing, to leverage these useful relations. Our method first models all the complex entities and relations in KT as a heterogeneous graph, and then uses a heterogeneous graph neural network to obtain entities' feature representations. After that, we feed the feature embeddings to a KT model in an end-to-end training manner. Due to the heterogeneous graph's high representational capacity, our method exploits the relations among students, questions, concepts, and answers in a concise and unified way. Experiments on four KT datasets show that our method achieves state-of-the-art performance.}, added-at = {2024-05-05T05:38:11.000+0200}, address = {Cham}, author = {Chen, Jisen and Shen, Jian and Long, Ting and Shen, Liping and Zhang, Weinan and Yu, Yong}, biburl = {https://www.bibsonomy.org/bibtex/2d61e1d288c175c3aae764d481e19ac3d/brusilovsky}, booktitle = {Neural Information Processing}, description = {Heterogeneous Graph Representation for Knowledge Tracing | SpringerLink}, editor = {Tanveer, Mohammad and Agarwal, Sonali and Ozawa, Seiichi and Ekbal, Asif and Jatowt, Adam}, interhash = {1f93b720d1c8f9dafcc2a70dc309ba8e}, intrahash = {d61e1d288c175c3aae764d481e19ac3d}, isbn = {978-3-031-30105-6}, keywords = {knowledge-tracing neural-network}, pages = {224--235}, publisher = {Springer International Publishing}, timestamp = {2024-05-05T05:38:11.000+0200}, title = {Heterogeneous Graph Representation for Knowledge Tracing}, year = 2023 }