[en] To obtain maximum data from the observation results some decomposition problems of data interpretation are considered. Algorithms intended to solve the nonlinear problems of data interpretation are given. Considered is the problem of identification and comparison of two spectra with a different number of components in them, arising in the course of evaluation of the model to experiment correspondence. It is shown that the approaches considered to the solution of decomposition problems of interpretation stage permit to computerize total mathematical treatment of observation results including a stable solution of inverse problems characteristic of spectrometric experiments