Abstract: We modeled pitch frequency by a sequential variational autoencoder to obtain the feature representations of the pitch accents of Japanese words for applications to Japanese speech education ...