Ahammed, Thaquib: Using Natural Language Processing to Automatically Score Student Responses In Virtual Lab Activities

Title: Using Natural Language Processing to Automatically Score Student Responses In Virtual Lab Activities

Name: Thaquib Ahammed

Major: Computer Science

School affiliation: School of Arts and Sciences

Programs: Aresty Summer Science Program

Other contributors: Janice Gobert

Abstract: Inq-ITS is a science-based classroom application that allows teachers to assign their students interactive virtual labs. Students conduct experiments where they manipulate different variables and see the results played out on their screen. While doing this, the student formulates hypotheses, observes data, and records their results. To show that they are comprehending not only the concepts of that specific lab, but also the practices of the scientific process, students provide written responses as well that detail what they learned from the lab. We have implemented a way that automatically scores these responses. We numerically categorize certain phrases that represent actions a student took or observed in the lab; like “I increased the height” or “the ball went faster”; and attach a point value to each of these phrases when they are found in the response. Other virtual lab applications on the market leave these written portions for the teacher to grade themselves. Not only is it more time- consuming for the teacher, but manual evaluation of these responses often does not reflect actual student comprehension. For instance, there could be a student that actually understands the lesson from the lab but is unable to properly articulate it. Therefore, a teacher may misunderstand this writing inability as the student not comprehending the topic. But because our application does not just focus on writing ability and places more emphasis on conceptual understanding, the automatic scoring we provide gives teachers a better interpretation of their student’s performance.