This chapter summarizes the work presented in this thesis. We have developed two first principles methods to calculate the dielectric function of condensed phases precisely and effectively. The first method employs the anharmonic phonon method and includes third and fourth-order anharmonicity in a self-consistent manner. The second method proposes the chemical bond based machine learning model for dipole moments, enabling the dipole moment prediction for various molecular systems and accelerating ab initio calculations. These methods demonstrate high accuracy when applied to real-world systems and reveal the importance of anharmonic effects or intermolecular interactions on the dielectric properties of materials. Finally, we discuss the future research directions.