Search for: All records

The use of writing-based exercises in a circuit analysis course has shown promise in aiding students likely to struggle in the course by enhancing their conceptual understanding of topics related to DC circuit analysis [1]. As grading of writing samples and providing personalized feedback can be time-intensive, automating the evaluation and feedback processes through use of emerging techniques in natural language processing (NLP) could open the door for more widespread use of such writing exercises across STEM courses, thus benefiting students in most need of assistance. In this paper, the development and initial testing of two web-based writing activities that leverage a basic NLP technique to probe student writing related to DC circuits are described. The first writing exercise has students describe what happens to the power of various elements in a resistive circuit as the value of one of the resistors decreases. The second exercise has students consider situations in which the ideal independent voltage and current source models might fail. Both writing exercises are built from a template that includes several metacognitive prompts to spur self-reflection on the part of the user. A rule-based approach was taken to detect evidence of common misconceptions [2] and errors in student responses, as well as to identify sentences that revealed the student was correctly addressing the problems. Based on identified misconceptions or correct concepts in a student’s writing, the web-based application selects appropriate directed line of reasoning (DLR) feedback paths to attempt to lead the writer to an accurate understanding of the behavior of the circuits in question. Key features of the web-based application template as well as details regarding misconception detection and personalized feedback are described. Student impressions of the value of the DLR feedback is assessed using comments provided by the student within the applications. Planned modifications of the web-based writing exercise template based on this formative assessment will be given and address a broader goal of this work – to develop a web-based template that instructors across STEM disciplines, even those without a background in coding, could use to implement their own conceptual writing exercises. 

Evidence suggests that bats are important hosts of filoviruses, yet the specific species involved remain largely unidentified. Niemann-Pick C1 (NPC1) is an essential entry receptor, with amino acid variations influencing viral susceptibility and species-specific tropism. Herein, we conducted combinatorial binding studies with seven filovirus glycoproteins (GPs) and NPC1 orthologs from 81 bat species. We found that GP-NPC1 binding correlated poorly with phylogeny. By integrating binding assays with machine learning, we identified genetic factors influencing virus-receptor-binding and predicted GP-NPC1-binding avidity for additional filoviruses and bats. Moreover, combining receptor-binding avidities with bat geographic distribution and the locations of previous Ebola outbreaks allowed us to rank bats by their potential as Ebola virus hosts. This study represents a comprehensive investigation of filovirus-receptor binding in bats (1,484 GP-NPC1 pairs, 11 filoviruses, and 135 bats) and describes a multidisciplinary approach to predict susceptible species and guide filovirus host surveillance.