Improving Feedback Recommendation in Intelligent Tutoring Systems using NLP Public
Downloadable Contentopen in viewer
The fundamental principle behind item-based recommender systems is to compare the similarity of an item to a ranked set of items. Our study explores the use of such recommendation system to suggest feedback comments for open-ended student answers in Intelligent Tutoring Systems. Several research works in the past have contributed to the development of technologies and strategies to assess and grade students’ work. The study in this paper aims to leverage Natural Language Processing, Machine Learning, and Statistical Methods to build a feedback infrastructure to help teachers in assessing their students’ open-ended answers. We investigate multiple approaches in determining the semantic similarity between two answers and evaluate the quality of semantic relatedness using our own metric called Teacher Agreement Score (TAS). It is often considered difficult task to assess open-ended data such as natural language. To evaluate the quality of the system, we have built a Software Infrastructure that enables running Randomized Control Trials on ASSISTments Platform to study the behavior by extracting information on the usage and effectiveness of the developed system.
- Defense date
- Date created
- Resource type
- Rights statement
Permanent link to this page: https://digital.wpi.edu/show/xd07gw642