TACCTI: Expediting Cultural Capital Identification in Student Reflective Journaling
By: Gian Carlo Baldonado
Department: Computer Science
Faculty Advisor: Dr. Anagha Kulkarni
TACCTI (pronounced as "tacit") is a computational framework introduced in the Alma project, which focuses on affirming the cultural wealth of historically underrepresented (HU) students in STEM through reflective journaling. The framework employs natural language processing and machine learning techniques to automatically identify instances of cultural capital themes (CCTs) in student reflective essays. TACCTI aims to provide a scalable and efficient solution to the labor-intensive task of identifying CCTs, such as attainment (describes a tangible goal) or first generation (identifies student being first in family college attendee). The first empirical evaluation of TACCTI using different machine learning algorithms shows that the framework achieves high accuracy in identifying the presence of CCTs in student essays, with F1 scores ranging from 0.81 to 0.92. Currently, TACCTI leverages a DistillBERT model that supports the identification of one CCT, attainment, but the framework is designed to facilitate expansion to additional CCTs. After a year hiatus, the project resumed in 2023 with a graduate student of the Data Science and AI Master’s program at SF State. Gian Carlo Baldonado, the graduate student lead of TACCTI, has developed a program that uses the existing DistilBERT model to label new student essays, which as of April 10, 2022, is able to annotate essays with a rate of 600 essays per minute. Gian aims to share new updates on TACCTI, including efforts on identifying CCTs from new data, error analysis of existing models, and expanding the capabilities of TACCTI to include more themes.