Locating disfluencies in spontaneous speech: an acoustical analysis