Measuring and Visualizing Metadiscursive Markers in Student Writing

Aneet Dharmavaram Narendranath, Michigan Technological University
Zachary Thelander
Sirena C. Hargrove-Leak, Elon University


Metadiscourse markers (MDM) are words or phrases that help connect and organize ideas or attitudes in genre-specific written discourse. Using rule-based or dictionary-based techniques, the presence of MDM may be confirmed and their distribution measured to help deconstruct genre-specific discourse. Such a deconstruction may help identify “something“about knowledge sharing, learning, topic or authorship attribution, and knowledge symmetries (or asymmetries) in cross-sectional or longitudinal analyses of writing. In this paper, we introduce a methodology consisting of data processing and visualization that is at the intersection of genre analysis, statistics, dimensionality reduction, and natural language processing. We apply this methodology to publicly available newsgroup data which is pre-labeled by topic to demonstrate that MDM distribution may be used to extract a visual dichotomy in the text structure belonging to different topics. In other words, text data pertaining to a specific topic have similar MDM distribution characteristics. In the future we will apply this methodology to labeled reflections authored by n students in an Engineering mechanics classroom that is infused with activities that involve an Entrepreneurial mindset (EM) to identify if MDM distribution and clustering indicates the presence of EM. Future work will also include exploring the confluence of MDM and rhetorical moves, since we believe this will support identifying EM, metacognition, and/or the achievement of a threshold skill. As part of this broader goal, we will create web-based digital tools to assess student writing and statistical regression models that would automatically classify the presence of an Entrepreneurial mindset in student writing.