Feedback is an essential component of learning and a key difficulty in achieving quality education at scale. Providing feedback is often a tedious task and there is a paucity of resources to aid teachers. In this work, we expand on previous tools that focus on generating natural language feedback for open response questions. Computer-based systems have the unique advantage of being able to collect action-by-action reports of the steps a student took to reach an answer along with metadata, such as time spent on a problem. It is difficult for teachers to analyze the detailed metadata when providing feedback, but it presents us with an opportunity to distill information from it. We take on problem-solving action data to provide teachers with detectors of student behavior. These detectors can be used to better keep track of their students' activity and inform what feedback can be provided.

Creator

• Alphonsus, Aaron G.

Contributors

• Heffernan, Neil

Degree

• MS

Unit

• Computer Science

Publisher

• Worcester Polytechnic Institute

Identifier

• etd-22686

Keyword

• lstm
• nlp
• behavior detection
• time series
• feedback

Advisor

• Heffernan, Neil

Orcid

• https://orcid.org/0000-0002-8081-0630

Defense date

• 2021-04-28

Year

• 2021

Date created

• 2021-05-06

Resource type

• Thesis

Rights statement

• In Copyright