The paper analyzes the accuracy of automatic segmental parsing with incorporated multiple pronunciation variants and describes a segmentation tool for Praat environment.