Personalized feedback for open-response mathematical questions using long short-term memory networks

Joshua J. Michalenko; Andrew S. Lan; Richard G. Baraniuk

Personalized feedback for open-response mathematical questions using long short-term memory networks

Joshua J. Michalenko, Andrew S. Lan, Richard G. Baraniuk

Research output: Contribution to conference › Paper › peer-review

Abstract

In this paper, we explore the problem of automatic grading and feedback generation for open-response mathematical questions. We resort to the long short-term memory (LSTM) network to learn the simple task of polynomial factorization and use the trained network for grading and feedback. We use Wolfram Alpha to synthetically generate a training dataset that consists of step-by-step responses to polynomial factorization questions to train the LSTM network. Preliminary results validate the efficacy of LSTMs in learning to factor low-order polynomials; we also demonstrate how to leverage the trained network for automatic grading and personalized feedback generation.

Original language	English (US)
Pages	350-351
Number of pages	2
State	Published - Jan 1 2017
Event	10th International Conference on Educational Data Mining, EDM 2017 - Wuhan, China Duration: Jun 25 2017 → Jun 28 2017

Other

Other	10th International Conference on Educational Data Mining, EDM 2017
Country/Territory	China
City	Wuhan
Period	6/25/17 → 6/28/17

Keywords

Automatic grading
Feedback generation
Long short-term memory networks
Mathematical expressions

ASJC Scopus subject areas

Computer Science Applications
Information Systems

Cite this

@conference{2c5f5cfd54db43aba8adeb70d59acb9a,

title = "Personalized feedback for open-response mathematical questions using long short-term memory networks",

abstract = "In this paper, we explore the problem of automatic grading and feedback generation for open-response mathematical questions. We resort to the long short-term memory (LSTM) network to learn the simple task of polynomial factorization and use the trained network for grading and feedback. We use Wolfram Alpha to synthetically generate a training dataset that consists of step-by-step responses to polynomial factorization questions to train the LSTM network. Preliminary results validate the efficacy of LSTMs in learning to factor low-order polynomials; we also demonstrate how to leverage the trained network for automatic grading and personalized feedback generation.",

keywords = "Automatic grading, Feedback generation, Long short-term memory networks, Mathematical expressions",

author = "Michalenko, {Joshua J.} and Lan, {Andrew S.} and Baraniuk, {Richard G.}",

year = "2017",

month = jan,

day = "1",

language = "English (US)",

pages = "350--351",

note = "10th International Conference on Educational Data Mining, EDM 2017 ; Conference date: 25-06-2017 Through 28-06-2017",

}

TY - CONF

T1 - Personalized feedback for open-response mathematical questions using long short-term memory networks

AU - Michalenko, Joshua J.

AU - Lan, Andrew S.

AU - Baraniuk, Richard G.

PY - 2017/1/1

Y1 - 2017/1/1

N2 - In this paper, we explore the problem of automatic grading and feedback generation for open-response mathematical questions. We resort to the long short-term memory (LSTM) network to learn the simple task of polynomial factorization and use the trained network for grading and feedback. We use Wolfram Alpha to synthetically generate a training dataset that consists of step-by-step responses to polynomial factorization questions to train the LSTM network. Preliminary results validate the efficacy of LSTMs in learning to factor low-order polynomials; we also demonstrate how to leverage the trained network for automatic grading and personalized feedback generation.

AB - In this paper, we explore the problem of automatic grading and feedback generation for open-response mathematical questions. We resort to the long short-term memory (LSTM) network to learn the simple task of polynomial factorization and use the trained network for grading and feedback. We use Wolfram Alpha to synthetically generate a training dataset that consists of step-by-step responses to polynomial factorization questions to train the LSTM network. Preliminary results validate the efficacy of LSTMs in learning to factor low-order polynomials; we also demonstrate how to leverage the trained network for automatic grading and personalized feedback generation.

KW - Automatic grading

KW - Feedback generation

KW - Long short-term memory networks

KW - Mathematical expressions

UR - http://www.scopus.com/inward/record.url?scp=85072299623&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072299623&partnerID=8YFLogxK

M3 - Paper

SP - 350

EP - 351

T2 - 10th International Conference on Educational Data Mining, EDM 2017

Y2 - 25 June 2017 through 28 June 2017

ER -

Personalized feedback for open-response mathematical questions using long short-term memory networks

Abstract

Other

Keywords

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this