Unsupervised Learning of Question Difficulty Levels Using Assessment Responses

Narayanan, Sankaran; Kommuri, Vamsi Sai; Subramanian, N.; Bijlani, Kamal; Nair, Nandu Chandran

doi:10.1007/978-3-319-62392-4_39

Cited by 6 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, such models need a large number of calibrated questions for training, which might be too costly to obtain, especially for smaller institutions. employs a pairwise difficulty comparison scheme similar to the one we will, but they still require the user responses for the algorithm to work, same as (Narayanan et al, 2017).…”

Section: Related Workmentioning

confidence: 99%

Towards the application of calibrated Transformers to the unsupervised estimation of question difficulty from text

Loginova¹,

Benedetto²,

Benoit³

et al. 2021

Proceedings of the Conference Recent Advances in Natural Language Processing - Deep Learning for Natural Language Processing Me

View full text Add to dashboard Cite

Section: Related Workmentioning

confidence: 99%

Towards the application of calibrated Transformers to the unsupervised estimation of question difficulty from text

Loginova¹,

Benedetto²,

Benoit³

et al. 2021

Proceedings of the Conference Recent Advances in Natural Language Processing - Deep Learning for Natural Language Processing Me

View full text Add to dashboard Cite

“…Specifically, the correct rate of the same student answering questions with similar difficulty or discrimination at the same knowledge concept is similar. Some existing work that study attributes of questions is primarily about mining them [5]. Other work try to use attributes of questions in knowledge tracing [6], but they do not specifically apply question attributes to deep knowledge tracing.…”

Section: Introductionmentioning

confidence: 99%

Attention-Based Knowledge Tracing with Heterogeneous Information Network Embedding

Zhang

Deng

et al. 2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Knowledge tracing is a key area of research contributing to personalized education. In recent times, deep knowledge tracing has achieved great success. However, the sparsity of students' practice data still limits the performance and application of knowledge tracing. An additional complication is that the contribution of the answer record to the current knowledge state is different at each time step. To solve these problems, we propose Attention-based Knowledge Tracing with Heterogeneous Information Network Embedding (AKTHE). First, we describe questions and their attributes with a heterogeneous information network and generate meaningful node embeddings. Second, we capture the relevance of historical data to the current state by using attention mechanism. Experimental results on four benchmark datasets verify the superiority of our method for knowledge tracing.

show abstract

“…The common method to obtain a valid and reliable DL is by collecting sufficient student performance data and report an Item Difficulty Level [10], [11] and a Discrimination Index [10]. Automated methods such as the accumulative test by Sokolova et al [12], the Question Classifier Engine by Narayanan et al [13], or the algorithm based on a Monte-Carlo approach by Sud et al [14] require actual student performance as input data. Therefore, it would be more efficient if we could rely on expert predictions of DL from.…”

Section: Background and Theoretical Frameworkmentioning

confidence: 99%

Comparing Expert Predictions to Student Performance on Challenging Conceptual Questions: Towards an Adaptive Learning Module for Materials Science

Nigon,

Simionescu,

Ekstedt

et al.

2022 ASEE Annual Conference &Amp; Exposition Proceedings

View full text Add to dashboard Cite

Thomas Ekstedt is a software developer in the School of Chemical, Biological and Environmental Engineering at Oregon State University. He is involved in the development of technology-based educational systems, particularly in the areas of concept-based instruction, adaptive learning, and interactive simulation of physical phenomena. Julie TuckerDr. Tucker earned her B.S. in Nuclear Engineering from the University of Missouri -Rolla. She attended graduate school at the University of Wisconsin -Madison as a Naval Nuclear Propulsion Fellow, where she received her M.S. and Ph.D. in Nuclear Engineering with a minor in Materials Science in 2008. After graduation, Dr. Tucker spent five years as a Principal Scientist at Knolls Atomic Power Laboratory in Schenectady, NY studying the thermal stability of structural alloys for nuclear power systems. She joined the School of Mechanical, Industrial, and Manufacturing Engineering at Oregon State University as an Assistant Professor in 2013 and was promoted to Associate Professor in 2019. In 2019 she was awarded the Dean's Professorship and was also appointed as the Materials Science Interdisciplinary Graduate Program Director. Dr. Tucker has an active research group focused on degradation of materials in extreme environments and alloy development. Her research efforts leverage both modeling and experimental approaches to gain fundamental understanding of materials performance. Milo KoretskyMilo Koretsky (he/him/his) is the McDonnell Family Bridge Professor holding a joint appointment in Chemical and Biological Engineering and Education at Tufts University. He received his BS and MS degrees from UC San Diego and his PhD from UC Berkeley, all in chemical engineering. He is interested in integrating technology into effective educational practices and in promoting the use of higher-level cognitive and social skills in engineering problem solving.

show abstract

Unsupervised Learning of Question Difficulty Levels Using Assessment Responses

Cited by 6 publications

References 9 publications

Towards the application of calibrated Transformers to the unsupervised estimation of question difficulty from text

Towards the application of calibrated Transformers to the unsupervised estimation of question difficulty from text

Attention-Based Knowledge Tracing with Heterogeneous Information Network Embedding

Comparing Expert Predictions to Student Performance on Challenging Conceptual Questions: Towards an Adaptive Learning Module for Materials Science

Contact Info

Product

Resources

About