Generating student feedback from time-series data using Reinforcement Learning

Gkatzia, Dimitra; Hastie, Helen; Janarthanam, Srinivasan; Lemon, Oliver

Abstract

We describe a statistical Natural LanguageGeneration (NLG) method for summarisa-tion of time-series data in the context offeedback generation for students. In thispaper, we initially present a method forcollecting time-series data from students(e.g. marks, lectures attended) and use ex-ample feedback from lecturers in a data-driven approach to content selection. Weshow a novel way of constructing a rewardfunction for our Reinforcement Learningagent that is informed by the lecturers’method of providing feedback. We eval-uate our system with undergraduate stu-dents by comparing it to three baselinesystems: a rule-based system, lecturer-constructed summaries and a Brute Forcesystem.Our evaluation shows that thefeedback generated by our learning agentis viewed by students to be as good as thefeedback from the lecturers. Our findingssuggest that the learning agent needs totake into account both the student and lec-turers’ preferences.

Citation

Gkatzia, D., Hastie, H., Janarthanam, S., & Lemon, O. (2013, August). Generating student feedback from time-series data using Reinforcement Learning. Presented at 14th European Workshop On Natural Language Generation, Sofia, Bulgaria

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 14th European Workshop On Natural Language Generation
• Start Date: Aug 8, 2013
• End Date: Aug 9, 2013
• Publication Date: 2013
• Deposit Date: Apr 21, 2020
• Publicly Available Date: Apr 21, 2020
• Publisher: Association for Computational Linguistics (ACL)
• Pages: 115-124
• Book Title: Proceedings of the 14th European Workshop on Natural Language Generation
• Public URL: http://researchrepository.napier.ac.uk/Output/1791775
• Publisher URL: https://www.aclweb.org/anthology/W13-2115.pdf