hi! (my name is Rebecca)

My research is motivated by a fundamental question:

‍how do representation learning paradigms shape the ways in which artistic and cultural expressions are created, consumed, and commodified?

This question has led me to explore:

recommender systems, particularly content discovery
multimodal representation learning, with an emphasis on the contextual semantics of music
algorithmic fairness, drawing on both machine learning, game theory, and human–AI interaction

Discovery

An interdisciplinary framework which embeds the algorithmic portions of content discovery in a broader ecosystem through which users engage with cultural expressions.
MusicSem is a dataset of over 32,000 language–audio pairs collected from organic online discussions, designed to capture the rich, contextual ways people naturally describe music and evaluate whether multimodal models can understand these nuanced musical semantics.
LARP is a multimodal representation learning framework that combines musical audio, textual descriptions, and playlist relationships through contrastive learning to improve music recommendation, particularly for newly released or lesser-known songs with limited listening histories (the cold-start problem).