Xinyi Xu
Pronouns: she/her
Research Mentor(s): Allyson Flaster
Research Mentor School/College/Department: Inter-university Consortium for Political and Social Research / Other
Program: UROP
Session: Session 5 (2:40pm – 3:30pm)
Authors: Annalise Paulson, Xinyi Xu
Abstract
Machine learning and natural language processing techniques have been one way for social scientists to analyze and understand complex social phenomena. We want to categorize different courses based on College Course Map (CCM) by using course numbers and course titles. However, because the model sometimes struggles using just course numbers and course titles, we are exploring methods to supplement it with the course description and other additional information to improve classification accuracy. In this context, this research aims to examine the method to get useful information from messy and unstructured course description data. We used supervised machine learning on text data, by using neural networks to classify sections of data into useful information about courses for social science research. To generate training data for this model, we scraped the course description from the course catalog, then annotate them with labels corresponding to useful pieces of information. This provides training data we use for supervised machine learning, training a model that can extract text from course descriptions at scale.
