Alternatives To Next Token Prediction In Text Generation -- A Survey