(paraphrasing would be a markov kernel here, and with the transitivity property I mentioned earlier, I’m asking that achieves its stationary distribution in one iteration)
for this condition, if you also want symmetricity, this is a very strong condition; you’d only accept “lossless paraphrasings”. i think not only are you achieving the stationary distribution in one iteration but the distribution cannot change, so this is either a markov kernel for every semantically different phrase, or not-markov.
for this condition, if you also want symmetricity, this is a very strong condition; you’d only accept “lossless paraphrasings”. i think not only are you achieving the stationary distribution in one iteration but the distribution cannot change, so this is either a markov kernel for every semantically different phrase, or not-markov.