Describing and predicting affective messages for expressive speech synthesis