Low-dimensional Embodied Semantics for Music and Language