Reproduce “Exploring Genre and Success Classification through Song Lyrics using DistilBERT: A Fun NLP Venture” 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
Servando Pizarro Martinez, Moritz Zimmermann, Miguel Serkan Offermann, Florian Reither
Researchers from the University of Cologne developed a system leveraging DistilBERT to classify song genres and predict both a song's success (based on views) and its approximate release year solely from lyrical content. The system achieved 65% accuracy for genre classification, 79% accuracy for success prediction, and an RMSE of 14.18 for release year estimation using extracted BERT embeddings with SVR.
Key insights
- DistilBERT effectively captures contextual semantic meaning in lyrics, outperforming traditional methods for complex classification tasks.
- Rap lyrics are highly distinguishable due to their unique vocabulary and stylistic elements, achieving an F1-score of 0.84 in classification.
- Even without fine-tuning a neural network, BERT embeddings combined with Support Vector Machine Regression yield competitive results for temporal prediction from lyrics.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.