Representations in vision and language converge in a shared, multidimensional space of perceived similarities