Non-globular domains in protein sequences: Automated segmentation using complexity measures