Proceedings of the 9th International Conference on Learning Analytics & Knowledge | 2019
Topic Development to Support Revision Feedback
Abstract
Revision is important but challenging for novice writers, particularly in post-secondary education where opportunities for personalized feedback are limited. Inexperienced writers typically overlook revision; when they do revise, they focus on surface errors rather than global revisions that enhance meaning and coherence. Writing analytics can automate personalized prompts to guide revision. We use topic modelling LDA as grounds for an analytic to scaffold holistic revision at paragraph and essay levels. The analytic visualizes topic distribution and generates three types of prompts: Introduction, Paragraph and Conclusion. Feedback encourages revisions focusing on sequencing topics, expanding underdeveloped ideas, and making holistic revisions to improve clarity and coherence of paragraphs. Model feedback was evaluated using undergraduate student essays on various topics scored by human evaluators. Model accuracy was strong for all types of feedback. This opens new branches of research to explore generating personalized feedback at paragraph and essay levels.