EN
Detecting and removing repetitions of semantically equivalent sentences improves the quality of the MML database, but it is also an important factor in evaluating the robustness of the Mizar system. However, the complicated structure of proofs in the MML database makes it very difficult to automatically discover and remove such repetitions. One possible solution to this problem may be to apply the delocalization process, i.e. moving some sentences to outer levels of a proof where the elimination of repetitions becomes possible.