What actually happens in aWriting Center? NLP and computational analysis applied to Writing Center data to create comprehensive topic models
Voiceover
Presenter(s)
David J DeBonis
CU Denver Undergrad Research Program
Undergraduate Research Opportunity Program (UROP)
Abstract or Description
A Writing Center (WC) collects tremendous amounts of data. Although the CU Denver WC consistently analyzes this data, this analysis is generally limited to discrete variables. However, there are also paragraphs of text-data collected for each session; this data is vital for pedagogy, but has not been applied to large-scale analyses. This study seeks to use Natural Language Processing (NLP) to better understand the WC’s data.
WC data from August 17, 2015 through December 09, 2020 was exported, then cleaned and analyzed using Python; this produced a dataset of 25,115 appointments. The analysis focuses on the following features from client report form (CRF) data: after each WC session, a consultant chooses up to three foci (from a predetermined list) which summarize the session; the consultant then describes the activity of the session through open-response questions.
Step one in this study’s NLP analysis was, for each appointment, to clean and concatenate CRF text-data into one text-set, then vectorize it by word-tokens. Step two was to apply TF-IDF scoring to statistically extract one feature word from each text-set. (The feature word is essentially the most salient word of the open-response, as determined by frequency/relative-proportion both within and across all text-sets.) For step three, these feature words were grouped by session focus and organized into topic models containing the top 20 feature words for each focus. Finally, models were organized into word clouds for visualization.
The value of analyzing the descriptive reflections of WC sessions is a systematic understanding of what happens during the learning experience of a WC consultation. In short, it provides both depth and breadth to the question, “What actually happens in a Writing Center?” This data in visual form may help students, faculty, and university administrators better conceptualize the value of WCs to student learning and success.