Difference between revisions of "Dynamic Knowledge Modeling in Textbook-Based Learning"
| Line 1: | Line 1: | ||
Various e-learning systems that provide electronic textbooks are gathering data on large numbers of student reading interactions. This data can potentially be used to model student knowledge acquisition. However, reading activity is often overlooked in canonical student modeling. Prior studies modeling learning from reading either estimate student knowledge at the end of all reading activities, or use quiz performance data with expert-crafted knowledge components (KCs). In this work, we demonstrate that the dynamic modeling of student knowledge is feasible and that automatic text analysis can be applied to save expert effort. We propose a data-driven approach for dynamic student modeling in textbook-based learning. We formulate the problem of modeling learning from reading as a reading- time prediction problem, reconstruct existing popular student models (such as Knowledge Tracing) and explore two automatic text analysis approaches (bag-of-words-based and latent semantic-based) to build the KC model. We evaluate the proposed framework using a dataset collected from a Human-Computer Interaction course. Results show that our approach for reading modeling is plausible; the pro- posed Knowledge Tracing-based student model reliably outperforms baselines and the latent semantic-based approach can be a promising way to construct a KC model. Serving as the first step to model dynamic knowledge in textbook- based learning, our framework can be applied to a broader context of open-corpus personalized learning. | Various e-learning systems that provide electronic textbooks are gathering data on large numbers of student reading interactions. This data can potentially be used to model student knowledge acquisition. However, reading activity is often overlooked in canonical student modeling. Prior studies modeling learning from reading either estimate student knowledge at the end of all reading activities, or use quiz performance data with expert-crafted knowledge components (KCs). In this work, we demonstrate that the dynamic modeling of student knowledge is feasible and that automatic text analysis can be applied to save expert effort. We propose a data-driven approach for dynamic student modeling in textbook-based learning. We formulate the problem of modeling learning from reading as a reading- time prediction problem, reconstruct existing popular student models (such as Knowledge Tracing) and explore two automatic text analysis approaches (bag-of-words-based and latent semantic-based) to build the KC model. We evaluate the proposed framework using a dataset collected from a Human-Computer Interaction course. Results show that our approach for reading modeling is plausible; the pro- posed Knowledge Tracing-based student model reliably outperforms baselines and the latent semantic-based approach can be a promising way to construct a KC model. Serving as the first step to model dynamic knowledge in textbook- based learning, our framework can be applied to a broader context of open-corpus personalized learning. | ||
| − | [[Image:ReadingLearningProcessExplain.png| | + | [[Image:ReadingLearningProcessExplain.png|300x300px]] |
[[Image:TrainExplain.png|400x400px]] | [[Image:TrainExplain.png|400x400px]] | ||
Revision as of 16:20, 4 December 2016
Various e-learning systems that provide electronic textbooks are gathering data on large numbers of student reading interactions. This data can potentially be used to model student knowledge acquisition. However, reading activity is often overlooked in canonical student modeling. Prior studies modeling learning from reading either estimate student knowledge at the end of all reading activities, or use quiz performance data with expert-crafted knowledge components (KCs). In this work, we demonstrate that the dynamic modeling of student knowledge is feasible and that automatic text analysis can be applied to save expert effort. We propose a data-driven approach for dynamic student modeling in textbook-based learning. We formulate the problem of modeling learning from reading as a reading- time prediction problem, reconstruct existing popular student models (such as Knowledge Tracing) and explore two automatic text analysis approaches (bag-of-words-based and latent semantic-based) to build the KC model. We evaluate the proposed framework using a dataset collected from a Human-Computer Interaction course. Results show that our approach for reading modeling is plausible; the pro- posed Knowledge Tracing-based student model reliably outperforms baselines and the latent semantic-based approach can be a promising way to construct a KC model. Serving as the first step to model dynamic knowledge in textbook- based learning, our framework can be applied to a broader context of open-corpus personalized learning.
Publications
- Huang, Yun, Michael Yudelson, Shuguang Han, Daqing He, and Peter Brusilovsky. "A Framework for Dynamic Knowledge Modeling in Textbook-Based Learning." In Proceedings of the 2016 Conference on User Modeling Adaptation and Personalization, pp. 141-150. ACM, 2016. (paper) (presentation)
