| Day 4 :D Re: "stochastic" flawless copy pasta from Google but a Markov chain is still an example of a (finite) state machine and is not itself an implementation of anything. ngram Language Models are an implementation though, and is not a competing concept: With a language model, you could talk about "a three word Markov chain" or you can simply say "a trigram". You can say "A Markov chain of variable length" or you can say "an ngram". That is all that is meant regarding those 2. If a Markov assumption is that you can predict the next word based on knowing all the previous words, then a bigram assumption would be that you can predict the next word based on the previous 1 word. A trigram assumption is that you can predict a word with 2 previous words, because they're all 3 part of the same trigram. More from Stanford on language models (LM): > "Models that assign probabilities to sequences of words are called language models or LMs. In this chapter we introduce the simplest model that assigns probabilities to sentences and sequences of words, the n-gram." > "Markov models are the class of probabilistic models that assume we can predict the probability of some future unit without looking too far into the past. We can generalize the bigram (which looks one word into the past) to the trigram (which looks two words into the past) and thus to the n-gram." https://web.stanford.edu/~jurafsky/slp3/old_jan23/3.pdf Looking forward to your next comment! |