Decoding identity-related expression in Chinese vocal music using computational acoustic and lyric features

Vocal performance provides a measurable interface between linguistic structure and musical expression, making it a tractable domain for investigating how culturally meaningful features are encoded in sound. In Chinese vocal music, this interface is particularly salient due to the interaction between tonal language systems and melodic organization, raising the question of how identity-related expression may be distributed across textual and acoustic dimensions. This Registered Report proposes a multimodal computational framework to examine identity-related expression in Chinese vocal music by integrating lyrics, speech, and singing data. A three-layer corpus will be constructed based on vocal works identified in recent Chinese core-journal literature on music and identity. Natural language processing (NLP) will be used to extract high-frequency and sentiment-bearing lyric units, which define preregistered anchor points for subsequent speech recording and audio segmentation. Acoustic features, including pitch (F0), intensity, and contour-based descriptors, will be extracted from both sung and spoken realizations of matched lyric units. Cross-modal relational features will be computed to quantify correspondence between speech and singing. This protocol specifies confirmatory analyses to examine how textual, acoustic, and cross-modal features relate to independently collected identity ratings, using correlation analysis, principal component analysis (PCA), and Random Forest Regression (RFR). Model performance will be evaluated via cross-validation, and feature importance will be used to compare the relative contributions of linguistic and musical dimensions. The hypotheses, preprocessing steps, and analytical procedures are specified in this Stage 1 protocol to ensure transparency and reproducibility. A pilot analysis conducted on a representative excerpt demonstrates that the proposed pipeline can reliably extract comparable acoustic features across modalities, supporting the feasibility of the preregistered design. By formalizing a computational pathway linking vocal performance features to identity-related expression, this study aims to contribute to performance science and computational musicology by providing a replicable framework for analyzing culturally embedded vocal expression.