Reproduce “Emotion Embedding Spaces for Matching Music to Stories” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Minz Won, Justin Salamon, Nicholas J. Bryan, Gautham J. Mysore, Xavier Serra
The authors map story text and music into a shared emotion space for music retrieval. DistilBERT encodes text, ResNet encodes audio, and a three-branch metric-learning model uses Word2Vec tags to connect text and music. On ISEAR queries with Word2Vec mapping, it achieves 0.4317 macro P@5 and 0.4953 macro MRR, exceeding the two-branch model, but not valence-arousal regression with manual mapping.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.