Fusing ASR Outputs in Joint Training for Speech Emotion Recognition | IEEE Conference Publication | IEEE Xplore