For offering proactive services (e.g., personalized exercise recommendation) to the students in computer supported intelligent education, one of the fundamental tasks is predicting student performance (e.g., scores) on future exercises. Unfortunately, the problem of extracting rich information existed in the materials (e.g., knowledge concepts, exercise content) of exercises to achieve both more precise prediction of student performance and more interpretable analysis of knowledge acquisition remains underexplored. To this end, we present a holistic study of student performance prediction. We first propose a general Exercise-Enhanced Recurrent Neural Network (EERNN) framework by exploring both student's exercising records and the text content of corresponding exercises, where we simply summarize each student's state into an integrated vector. For making final predictions, we design two implementations on the basis of EERNN with different strategies, i.e., EERNNM with Markov property and EERNNA with Attention mechanism. Then, to explicitly track student's knowledge acquisition on multiple knowledge concepts, we extend EERNN to an explainable Exercise-aware Knowledge Tracing (EKT) framework by incorporating the knowledge concept information, where the student's state vector is extended to a knowledge state matrix. The experimental results on a large-scale data demonstrate the effectiveness of two frameworks as well as the interpretability of EKT.

Yiming Cui, Ting Liu, Wanxiang Che, Li Xiao, Zhipeng Chen, Wentao Ma, Shijin Wang, Guoping Hu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

This paper considers the direction of arrival (DOA) estimation of coherent and uncorrelated targets by exploiting a flexible multiple input multiple output (MIMO) radar. First, the proposed radar is generalized with two coprime expansion factors for enlarging the inter-element spacing of transmit and receive arrays, referred to as sparse arrays with flexible inter-element spacing (SA-FIS), which shows that the conventional nested or coprime ones are the special cases. The range of consecutive lags and the number of unique lags in the sum-difference coarray are derived in closed-form. It is verified that SA-FIS can obtain the maximum unique lags and also suppress the mutual coupling effects. We then extend the FIS viewpoint to sparse nonuniform linear arrays, where the two-level nested and coprime arrays are employed to achieve a significant increase in degrees of freedom (DOFs). Furthermore, to fully utilize these unique lags, we propose a reduced-complexity two-step sparse representation (SR) algorithm. By modifying and removing the off-diagonal elements of the estimated target covariance matrix, the proposed method can just identify the diagonal ones, thereby leading to further improved performance with much lower computational complexity. Finally, the Cramer-Rao bound (CRB) on DOAs and correlated coefficients for SA-FIS is derived. Numerical simulations demonstrate the superiority of the proposed method with SA-FIS in terms of DOFs, computations, estimation accuracy and resolution capability compared with previous ones.

Xiyuan Yang, Xiaotao Gu, Sheng Lin, Siliang Tang, Yueting Zhuang, Fei Wu, Zhigang Chen, Guoping Hu, Xiang Ren. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Yiming Cui, Wanxiang Che, Ting Liu, Bing Qin, Shijin Wang, Guoping Hu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.