For efficiently calibrating the correlation structure of a LIBOR market model (LMM) to market data, low-rank correlation parameterizations are necessary. In this paper we present a new simple approach for generating low-rank low-parametric forms from given full-rank parameterizations