Training Audio Captioning Models without Audio | IEEE Conference Publication | IEEE Xplore