Altoe, F. & Joyner, D. A. (2019). Annotation-free Automatic Examination Essay Feedback Generation. In 2019 IEEE Learning With MOOCS (LWMOOCS). IEEE.
Abstract: Many of the demands of scale appear at odds with quality, such as in the case of online MOOC degree programs. This paper presents research that focuses on reducing this dichotomy. Examination essays are important learning tools in for-credit courses. Psychology studies show they drive deeper learning strategies and invoke higher levels of cognitive processing in students. High-quality feedback is another critical aspect of students’ learning in advanced education. Key components of high-quality feedback are how expeditiously feedback is presented and how unambiguous it is. Examination essays are time-consuming for educators to grade, and at scale this may create an incentive for superficial feedback that may also reach students in a less than ideal turnaround time. This imposes clear scalability constraints and a potential decrease in course quality in online for-credit MOOC programs. We propose an annotation-free artificial intelligence-based approach for the automatic generation of examination essay rubrics and its subsequent utilization as part of high-quality feedback to students. It utilizes a combination of the natural language processing techniques TextRank and semantic similarity for automatic rubric generation and concept map creation for feedback presentation. The feedback is immediately available upon essay submission and offers the student a conceptual analysis of the submitted essay against the assignment’s learning objectives.