Meryem Yilmaz Soylu and Adrian Gallard, a research scientist and digital learning data analyst at C21U, presented a talk titled "Using NLP to Identify Leadership Skills for Online Master's Program Admissions" at the Affordable Degrees at Scale Symposium on December 5, 2023. The presentation delved into the significance of leadership skills in graduate education and their crucial role in admissions.
The project started with the team collaborating with Georgia Tech's Online Master of Science in Analytics program to extract a dataset of letters of recommendation (LORs) to train machine learning models. During the talk, Dr. Yilmaz Soylu detailed the process of creating a library of phrases and keywords indicating leadership and its sub-components, which served as a guide for human annotators. She highlighted the efficiency of data-efficient methods in producing robust models with minimal annotation. Dr. Yilmaz Soylu also mentioned using BERT as the primary model and the team's ongoing efforts to utilize alternative models like Gan-BERT, RoBERTa and SetFit, and that they validated the results at each phase through comparisons with test datasets and separate annotations.
Dr. Yilmaz Soylu said, "Our ultimate goal is to streamline and enhance the admission process by leveraging AI to evaluate and identify leadership skills. Additionally, our tool has immense potential to serve as a valuable companion throughout the online graduate education journey, keeping a record of students' progress and providing feedback and resources for further improvements on leadership and other 21st-century skills."
To make the AI tool outcomes accessible, they are concurrently developing a dashboard and user interface.