Zero-Shot Learning and Support Vector Machine Methods for Automated Abstract Screening in Scoping Reviews

• In Event: Leveraging Artificial Intelligence and Automation Tools for Systematic Reviewing

Abstract

Evidence-based literature reviews such as systematic reviews, meta-analyses, and scope reviews are widely used to synthesize research studies across disciplines. These reviews provide evidence-based information related to a specific topic that informs policy, practice, or theory development. The process of selecting, evaluating, and synthesizing relevant studies is labor-intensive and time-consuming. The advancement of machine learning and text-mining algorithms has facilitated the automation of several steps in the review process. This paper applies a hybrid method that combines the zero-shot learning (ZSL) method, Support Vector Machine (SVM), and manual validation to a sample of abstract screening in preparation for a scoping review on machine learning methods in education. It examines the strengths and limitations of this method and its application.

Authors

• Moe Debbagh Greene, Virginia Commonwealth University

• Michael Broda, Virginia Commonwealth University

• Chin-Chih Chen, Virginia Commonwealth University

• Chi-Ning Chang, Virginia Commonwealth University

• Tzu-Wei Wang, Virginia Commonwealth University

• Jeen Mariam Joy, Virginia Commonwealth University

• Yuyan Xia, University of Kentucky