Document clustering via dirichlet process mixture model with feature selection