Interpretable embeddings to understand computing careers

Abstract

We propose an approach for analyzing and comparing curricula of study programs in higher education. Pre-trained word embeddings are fine-tuned in a study program classification task, where each curriculum is represented by the names and content of its courses. By combining metric learning with a novel course-guided attention mechanism, our method obtains more accurate curriculum representations than strong baselines. Experiments on a new dataset containing curricula of computing programs demonstrate the interpretability power of our approach via attention weights, topic modeling, and embeddings visualizations. We also present a use case that compares computing study programs in the US and Latin America and showcase the capabilities of our method for identifying similarities and differences in topics of study in curricula from different countries.

Tasks

Reproductions