Short-answer grading using textual entailment

Issue title: Special Section: Intelligent and Fuzzy Systems applied to Language & Knowledge Engineering

Guest editors: David Pinto and Vivek Singh

Article type: Research Article

Authors: Basak, Rohini^a | Naskar, Sudip Kumar^a | Gelbukh, Alexander^{b; *}

Affiliations: [a] Jadavpur University, Kolkata, India | [b] CIC, Instituto Polytéchnico Nacional, Mexico City, Mexico

Correspondence: [*] Corresponding author. Alexander Gelbukh, CIC, Instituto Polytéchnico Nacional, Mexico City, 07738, Mexico. E-mail: gelbukh@gelbukh.com.

Abstract: Given a question, a reference answer, and the answer given by the student, the aim of the automatic short answer grading task is to assign a grade to the student’s answer. We use for this a large number of matching rules relying on recognizing entailment relation between dependency structures of the two answers. Comparison of the grades generated by our method with those given by human judges on a computer science dataset shows a quite promising maximum correlation of 0.627.

Keywords: Automatic short answer grading, recognizing textual entailment, dependency parsing, semantic similarity

DOI: 10.3233/JIFS-179038

Journal: Journal of Intelligent & Fuzzy Systems, vol. 36, no. 5, pp. 4909-4919, 2019