Using forced alignment for automatic acoustic-phonetic segmentation of aphasic discourse.