Contrastive Learning-based Audio to Lyrics Alignment for Multiple Languages
Simon Durand, Daniel Stoller, Sebastian Ewert
Disentangled sequential autoencoders (DSAEs) represent a class of probabilistic graphical models that describes an observed sequence with dynamic latent variables and a static latent variable. The former encode information at a frame rate identical to the observation, while the latter globally governs the entire sequence. This introduces an inductive bias and facilitates unsupervised disentanglement of the underlying local and global factors. In this paper, we show that the vanilla DSAE suffers from being sensitive to the choice of model architecture and capacity of the dynamic latent variables, and is prone to collapse the static latent variable. As a countermeasure, we propose TS-DSAE, a two-stage training framework that frst learns sequence-level prior distributions, which are subsequently employed to regularise the model and facilitate auxiliary objectives to promote disentanglement. The proposed framework is fully unsupervised and robust against the global factor collapse problem across a wide range of model confgurations. It also avoids typical solutions such as adversarial training which usually involves laborious parameter tuning, and domainspecifc data augmentation. We conduct quantitative and qualitative evaluations to demonstrate its robustness in terms of disentanglement on both artifcial and real-world music audio datasets.
Simon Durand, Daniel Stoller, Sebastian Ewert
Jakob Zeitler, Athanasios Vlontzos, Ciarán Mark Gilligan-Lee
Graham Van Goffrier, Lucas Maystre, Ciarán Mark Gilligan-Lee