Social work in the classroom? A tool to evaluate topical relevance in student writing

Heeryung Choi, Zijian Wang, Christopher Brooks, Kevyn Collins-Thompson, Beth Glover Reed, Dale Fitch

Abstract: In a climate where higher education institutions are actively aiming to increase inclusivity, we explore how a deep learning-based tool focused on text analysis is able to help assess how students think about issues of privilege, oppression, diversity and social justice (PODS). We created a vocabulary boosting and matching tool augmented with domain-specific corpora and relevance information. We find that the adoption of domain-specific corpora enhances model performance when identifying PODS-related words in short student-written responses to writing prompts, by building a more highly focused PODS vocabulary.