Reproduce “Vocal Music under Phoneme-Conditional Analysis” 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
Hayoon Kim, Kyogu Lee
Phoneme-conditional analysis measures how language-specific sound units shape vocal performance while controlling for singer, melody, genre, and recording context. Across 5,024 songs in nine languages, phoneme-level contrasts were especially visible in timbre and voice quality, and their combined profiles enabled 85.5% balanced-accuracy language classification, although many speech-based duration and pitch contrasts were weakened or reshaped in singing.
Key insights
- Phonological contrasts remain detectable in singing, but their acoustic realization is channel-dependent: timbre, spectral tilt, and voice quality generally show stronger effects than pitch or duration.
- Speech contrasts are often attenuated in singing; some duration patterns are shortened or even reversed, indicating that musical timing can reshape phonetic realization rather than simply preserve it.
- Language profiles are driven more by phoneme-conditional contrast features than by broad phoneme distributions, but classification alone does not prove that larger local effects directly accumulate into language-level vocal conventions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.