A Step Toward an Automatic Handwritten Homework Grading System for Mathematics

Journal article

Authors/Editors

NATASHA DEJDUMRONG

Strategic Research Themes

Machine Learning (Big Data Analytics)

Publication Details

Author list: Chaowicharat, Ekawat; Dejdumrong, Natasha;

Publication year: 2023

Volume number: 52

Issue number: 1

Start page: 169

End page: 184

Number of pages: 16

ISSN: 1392124X

URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85152682815&doi=10.5755%2fj01.itc.52.1.32066&partnerID=40&md5=885c727fff8cb83ec75b40f6b5562eb9

Languages: English-Great Britain (EN-GB)

View in Web of Science | View on publisher site | View citing articles in Web of Science

Abstract

An automatic system that helps teachers and students verify the correctness of handwritten derivation in mathematics homework is proposed. The system acquires input image containing handwritten mathematical deri-vation. In our preliminary study, the system that comprises only mathematical expression recognition (MER) and computer algebra system (CAS) did not perform well due to high misrecognition rate. Therefore, our study focuses on fixing the misrecognized symbols by using symbols replacement and the surrounding information. If all the original mathematical expressions (MEs) in the derivation sequence are already equivalent, the derivation is marked as “correct”. Otherwise, the symbols with low recognition confidence will be replaced by other possible candidates to maximize the number of equivalent MEs in that derivation. If there is none of symbols replacement that makes every line equivalent, the derivation is marked as “incorrect”. The recursive expression tree comparison was applied to report the types of mistake for those problems marked as incorrect. Finally, the performance of the system was evaluated by the digitally generated dataset of 6,000 handwritten mathematical derivations. The results showed that the symbols replacement improve the F1-score of derivation step marking from 69.41 to 95.95 % for the addition/ subtraction dataset and from 61.45 to 89.95 % for the multiplication dataset when compared to the case of using raw recognized string without symbols replacement. © 2023, Kauno Technologijos Universitetas. All rights reserved.

Keywords

automatic homework grading, Mathematical expression recognition, symbols replacement