Exploring Interpretable Latent Structure in Modern Data by Bayesian Modeling: Theory and Applications